Skip to content
Agents tracked: 492 Releases (7d): 545 Agent status: 1 with issuesSince last limit reset: Claude 9d · Codex 7d For AI agents: Markdown Updated Nov 20, 2026

AI agent releases: Sep 14 – Sep 20, 2026

299 stable releases from 107 tracked agents and frameworks in the week of Sep 14, 2026, newest first, with the headline changes from each set of release notes. Our lists keep each project's latest releases, so a very active project's older releases can be missing.

Sep 20, 2026

  • phone-harness cloud: rent a phone from the CLI; Android helpers work as promised
  • cloud start opens the phone's live view
  • android: a never-idle screen is asked once, then read from pixels
  • feat(composer): restore compact branch picker UX
  • feat(composer): choose a base ref in the New Workspace composer
  • feat(editor): add opt-in collapsed unchanged regions for file diffs
  • fix(android): keep the composer's corners fixed as the draft grows
  • fix(computers): give every engine a provider-safe desktop tools/list
  • fix(openai-compat): request streaming usage totals
  • Pi and OMP history: track local tokens and estimated costs across the app, CLI, and widgets without double-counting Claude or Codex usage.
  • More complete spending displays: add DeepSeek and OpenRouter widget balances and provider-reported spending in the web dashboard.
  • More reliable refreshes: correct paginated Codex costs, recover transient credential-access failures, and prevent crashes from oversized provider values.
  • Security: the svg/all extras now floor Pillow at >=12.3.0 for CVE-2026-54058 (Pillow was pulled in transitively via matplotlib).
  • Feature: a Go interface's method requirements (type Foo interface { Bar() }) now attach to the interface node, so calls resolve to them
  • Feature: a Swift protocol's method requirements (protocol P { func f() }) now attach to the protocol node
Open Connector
  • feat: support selective provider bundles in headless Bun builds
  • fix(feishu): list_drive_permissions cannot be called, and returns empty when it can
  • feat(slack): cursor paging, history windowing, credential identity, and two response bugs
  • Stored values resolve at run time — a step that stores a parameter now records what this run actually produced, including an empty string, instead of falling back to the author-time value.
  • Replayed API and JavaScript steps feed the next step — the response from a replayed API call and the result of a replayed JS step reach the operation that depends on them, so downstream steps see…
  • JS results are recorded as JSON — no more Python-style reprs in results, so values are directly usable and comparable.
  • commit: 18a3e3d03b1b94296cec7f1881e6dcbca7428a85
  • source tag: beta2026-09-20_18-13-27
  • regions: cn global
  • Meta Muse provider — Sign in with Meta using /login meta or use META_API_KEY to access Muse Spark models. See Meta (Muse subscription).
  • Added Meta (Muse subscription) login via /login meta with automatic Model API key refresh, plus META_API_KEY support
  • Enabled Node's persistent compile cache before loading the bundled CLI runtime, reducing repeat launch time.
  • Jev browser handoff now handles whole multi-step tasks, continuing through uncertain pages and returning to the parent only when finished or blocked
  • Memory recall uses Jev to select relevant project and global memories without loading a local embedding model at startup
  • Save Anthropic prompt-cache retention preferences with /cache extend and see clearer cache usage, retention estimates, and provider-specific limits
  • deps(website): bump astro to 7.2.8 (GHSA-26w7-cxv4-gfx2)
  • feat(analytics): separate basic usage from consented insights
  • docs(agent): keep repository guidance contextual
  • 加强存储、队列、路径锁和文件系统稳定性:修复 cp/mv 索引读取、空白资源导入、文件目标误作目录、PathLock 接管、缺失目标锁定、QueueFS 后台队列隔离、completion 状态归属和本地向量存储进程锁。
  • 扩展 Agent 和记忆接入能力:新增 Hermes、MiMo/MiMoCode、WorkBuddy 日志源,新增独立 Hermes OpenViking memory provider,支持私有网关 header,并改进插件 hook 与 MCP proxy 连接管理。
  • 改进检索和 MCP 行为:新增远程 VikingDB glob,修复 MCP grep 错误报告,MCP search 会拒绝 context-only 参数在 list 模式下被静默忽略。
  • OpenCode Go 订阅供应商:填入订阅的 API Key,即可使用 GLM、Kimi、MiniMax、Qwen、Grok、GPT 等模型。不同模型的协议差异由本地代理自动分流,无需手动配置。
  • Composer 能力菜单:输入框的「+」菜单现在可搜索,Skills、连接器、Agents、工作流、Agent Teams 和 Computer Use 都从这里进入。
  • 批准计划时切换执行模型:在执行计划确认条上直接选模型——同供应商立即切换,跨供应商自动重启并继续执行,不用重新描述需求。
  • feat(cli): simplify option names and add dump aliases
  • fix(service-worker): reject async module evaluation graphs
  • fix(modules): preserve constructors through module trees
No release notes published.
  • Reach sequencing runs by accession. New ENA tools resolve a public ENA/INSDC study, experiment, sample, or run accession to its sequencing runs — returning study, sample, experiment, organism…
  • Gene-set enrichment powered by g:Profiler. The Genes connector can now run GO and pathway enrichment for a differential or marker gene set, with organism scoping, selectable evidence sources, and…
  • NCBI reference genome lookups. Standalone NCBI tools resolve taxon names with ambiguity reporting, inspect versioned genome assemblies, and look up sequence aliases — historical assembly…
  • fix(studio): serve composition material in the ranges a media element asks for
  • feat(beatapi): serve the installed video and image models on BeatAPI
No release notes published.
  • DeepSeek now offers DeepSeek Flash and DeepSeek V4 Pro as easy provider
  • DeepSeek Flash supports image input and a 1M-token context window.
  • Fixed the provider model picker loading state so available choices load
  • chore(deps): bump anyio in the uv group across 1 directory (611614948) — dependabot[bot]
  • fix(state_store): long App.update() runs on Linux no longer hang permanently — an LMDB write transaction could resume on a different runtime thread and leave the writer lock held forever, blocking…
  • Fixes for duplicate execution of memoized components under concurrency
  • Concurrent runs of a memoized component with the same memo key (e.g. two overlapping App.update() calls on an app whose main function is memoized) now execute the body once: the queued run reuses…

Sep 19, 2026

  • Prompt cache warming — Keep valuable prompt caches alive during long tool runs and optionally while idle using cost-aware refreshes. See Cache Warming.
  • Bug reporting — Report problems with /bug using redacted diagnostics, optional transcripts, or exported ZIP archives. See Reporting Bugs.
  • Transcript-aware prompt and tool updates — Preserve instruction and tool changes across resume and branch navigation while retaining cached prefixes. See before_agent_start.
Hypit
  • fix(media-execution): preserve AAC tail for non-integer frame rates
  • fix: treat snapshot HTML URLs like capture and reject invalid --studio bases
  • fix(media-execution): preserve still frame domains across frame rates
  • fix(supervision): genuine-operator gating + completed-assistant progress semantics
  • fix(tui): clear inactive sidebar agents on subagent completion and idle events
  • fix(task-session-manager): gate all terminal publications on runtime evidence
  • History that follows your quotas: compare Codex and Claude costs and tokens across current and recent weekly quota windows, with history preserved across restarts.
  • An Overview that fits your menu: choose a compact layout, hide individual provider details, and share the Usage & Spend summary you see.
  • Clearer widgets for multiple accounts: pin saved accounts, see the quota closest to its limit, and retain full provider names and secondary allowances.
Planning With Files
  • Two maintainer-found bugs, fixes #275 and #274.
  • PowerShell route on OneDrive. resolve-plan-dir.ps1, the Cursor hook resolve-plan-context.ps1 and set-active-plan.ps1 refused an .active_plan pointer that carried the ReparsePoint file attribute.
  • Test guard walked the whole checkout. tests/test_session_catchup_copies_exact_basename.py loaded any session-catchup.py it found under the repository root, tracked or not.
  • Version 0.2.3
  • A waveform's bars are one pitch on screen at every zoom
  • Create ARCHITECTURE-TODO.md
  • Windows: .exe installer or portable. macOS: .dmg (Apple Silicon aarch64 or Intel x86_64). Linux: .AppImage (universal) or .deb (Debian/Ubuntu).
  • Debian 12+ ships without FUSE 2 by default, which AppImage requires. If ./omniget.AppImage fails with a libfuse error, run
  • bash sudo apt install libfuse2
  • ab56581f40471789539593c3f773b921e0b99da6 feat: add RADIUS Message-Authenticator support with a per-provider switch
  • 72a1d5a90fe1c9128901e1b946514dd865863250 feat: add organization-level exclusive signin across applications
  • 0352f4beaa5049b0037af145a2205ef51d433c67 feat: render the forced password update as a standalone auth page
  • docs: weekly changelog 2026-09-13 (pull request pane, subagents, Muse Code and Devin)
  • fix(host-service): classify a push refused by a hook or the remote as a non-500
  • chore(desktop): bump version to 1.29.0
  • Add the optional animated Tur Tur Sahur desktop companion.
  • Show normal and unattributed reload warnings only in the last 30 seconds; retain the five-minute Pro warning.
  • Improve automatic Continue, queued-message recovery, native Stop handling and recorded turn ownership.
  • #3922 9df7a9c Thanks @sailist! - Fix new messages occasionally landing at an old position in the conversation after resuming a session.
  • #3911 88a7d93 Thanks @Grapedge! - Fix compaction failing after switching to a model with a smaller context window.
  • #3929 6ffdf0d Thanks @7Hanrui! - The agent no longer assumes the current working directory is the project root.
  • feat(viewer): export a session to a self-contained HTML file
  • fix(allowlist): exclude pytest-style test_*.py from review
  • chore: drop the last TEMPORARY english-only exemption
  • [OPIK-8386] [BE] [FE] fix: govern Claude sampling params by the model, not the provider
  • [OPIK-8376] [BE] feat: fail readiness when the spans wrap flag disagrees with the DB topology
  • [OPIK-8366] [BE] feat: spans cutover drivers, runbook and rehearsal harness
  • Let TypeSafeModel fill a tool's arguments when Jev can express them
  • Add RealtimeSession.wait_for_playback() and let the docs examples finish the reply before closing
  • Add supports_text_output to ModelProfile, and run LLMJudge and GEval on a model without it
  • Changed auto mode for Claude API and Enterprise users, and on Bedrock, Vertex, Foundry and gateways, to default to the server-side classifier, which does not charge for classifier overhead…
  • Added an Auto mode server row to /status showing whether this session's auto mode classifier runs on the server
  • Configure root-model reasoning effort independently for swarm and deep swarm, with the selected effort shown in the effort switcher
  • Delegate multi-step browser tasks to a bounded Jev browser session that returns control to the parent agent
  • Create and manage separate panels from Markdown or linked PDFs, with bounded PDF payloads and safe Markdown fallback for older TUI clients
  • Release notes — formatted for people, with expandable sections.
  • Changelog — plain Markdown for AI agents and tools.
  • npm package
  • 飞书 / 钉钉 / 企业微信 OAuth SSO 登录与绑定
  • 可选登录验证码(滑块及多家云验证码),并提供 octop captcha reset
  • 容器工作区默认、专家头像,以及聊天 HITL / 技能选择体验优化
  • This release brings automatic Learning skill delivery to BuiltInAgent, more resilient SSE streaming, and richer runtime telemetry, alongside a set of targeted fixes across core, react-core, and Vue.
  • fix(test): isolate repair worker watchdog (ae00def77) — Anwesh
  • feat(cli): attach context files to opensre ask (25d0b5161) — Anwesh
  • ci(analytics): probe telemetry identity across operating systems and containers (730ebc1ad) — Vincent Hus
  • Added inline Kitty graphics rendering for LaTeX math equations ($$...$$ and fenced math, latex, and tex blocks) and Mermaid flowcharts and sequence diagrams in the artifact detail viewer, with…
  • Improved the ask_question interactive prompt with Left/Right arrow navigation between questions, inline previews of confirmed write-in answers, check marks on answered options, and single-step…
  • Improved the CLI startup banner for enterprise accounts to display the active Google Cloud Project ID beneath the signed-in account and plan tier
  • Agent swarms arrive in beta: durable pools of subagents with scoped messaging, inline progress badges, and Modal-backed workers. Deployments can turn swarms off.
  • Connecting Slack got much easier, with provider-managed installation, a guided bot onboarding checklist, and sturdier thread and DM conversations.
  • You can steer a running task with files and images, conversation documents now go straight into model inputs, and suggested prompts are collaborative and contextual.

Sep 18, 2026

Hypit
  • feat(pixverse): add the PixVerse V6 model package
  • fix(studio,media): secure local mutations and avoid blocking video fetch
  • fix(cli,runtime): UTF-8 secret input and recoverable worker startup
  • Save drafts. The new Drafts plugin is bundled and enabled by default. Save a message from a new-thread or follow-up composer, then choose Send now when you’re ready.
  • Redesigned handoffs. Hand off from the follow-up composer while preserving your draft and workspace.
  • Control finished turns. Choose whether finished turns collapse for each provider in Settings → Providers. Claude Code now keeps finished turns expanded by default.
  • Keyless-first experience: local model downloads are announced and many flows now run cleanly with no LLM key.
  • Pipeline robustness: pipeline runs are stamped at start and recover reliably after crashes, preserving completed documents.
  • Model wiring: image and audio transcription models are passed through end-to-end and the embedding model is recorded per dataset to ensure consistent embeddings.
  • Feature: Terraform block attributes (ami, instance_type, cidr_block, tags, and the like) are now preserved on the resource/data/module node and are queryable and searchable, with typed values…
  • Feature: JavaScript inside an inline <script> block of a PHP file is now indexed as JS (functions and calls) under the PHP file node, mirroring the Vue/Svelte embedded-script handling, with source…
  • Feature: a Rust self.method() call now resolves across files for simple generic impls (impl<T> Foo<T>), extending the split-impl resolution to generic types while staying fail-closed on bounded…
Deep Agents
  • Scoped cancel() requests to the requested session.
  • dfed90c release(deepagents-acp): 0.0.12
  • fc6df1f chore(deps): bump anyio from 4.12.1 to 4.14.2 in /libs/acp
  • .NET: fix: race concdition in workflow formula state
  • .NET: update remaining dependencies
  • .NET: fix: inspect for delemiter value before adding headers
  • Telemetry: OpenTelemetry event names are now preserved on platforms outside of Agent Engine, ensuring more accurate and consistent tracking. (9110770)
  • telemetry: only drop the OTel event name on Agent Engine (9110770)
Claude Code
  • Added AGENTS.md support: in a project with no CLAUDE.md, Claude Code reads AGENTS.md instead; change it under "Project instructions" in /config (not yet on Bedrock, Vertex or Foundry)
  • Added CLAUDE_GATEWAY_PROXY_IS_EGRESS_BOUNDARY=1 for Claude apps gateways whose only egress is a forward proxy: every outbound request hands the proxy the hostname instead of resolving it locally
  • Added an optional headers: map on Claude apps gateway upstreams, to send static headers to a proxy you run in front of a provider
  • Five new providers: track Nous Portal and Muse Code subscriptions, CodeRabbit reviews, Replicate spend, and Hugging Face Inference Providers charges and ZeroGPU quota.
  • More complete allowances: see Mistral API and Vibe Code allowances, Venice subscription credits, and Grok usage-limit reset coupons.
  • Richer spending details: view DeepSeek spend per model, OpenRouter pay-as-you-go spending and account Activity, and provider history in CLI output.
  • Changes since langchain==1.4.1
  • release(langchain): 1.4.2 fix(langchain): preserve model-generated tool calls in HITL tool call edits and add notice to ToolMessage
  • Stream large saved scan JSON output
  • Support custom patch validation prompts
  • Include Terraform files in scan inventories
  • init-session.sh: a plan name with an embedded newline created an unusable .planning/<date>-line-one<newline>line-two/ directory and exited 1 with it left behind.
  • Both initializers bound PLAN_ID for the attestation call but not PWF_PLAN_ROOT, so a pin inherited from another project silently redirected or blocked the attestation of the plan that was just…
  • session-catchup.py, OpenCode adapter: a part row whose state or state.input is not an object raised AttributeError, and a row whose data column is not valid JSON made SQLite raise malformed JSON…
  • feat(viewer): restyle the review comments section
  • feat(viewer): restyle session detail header and meta bar
  • feat(viewer): restyle the sessions list page
  • [OPIK-3268] [BE][FE] feat: let users name experiments from the Playground
  • [OPIK-8274] [SDK] perf: stream experiment bulk uploads instead of four full passes
  • #3878 f233f9d Thanks @7Sageer! - Stop listing unavailable media tools in subagent menus and clarify Read errors for images and videos.
  • #3762 c5ad17f Thanks @7Sageer! - Providers can read their API key from a named environment variable (api_key_env in config.toml or the provider API).
  • #3896 73ebe9a Thanks @sailist! - Keep follow-up messages sent during a turn from appearing twice after a session reload.
  • #2249 ebab868 Thanks @mattzcarey! - WebSockets speaks the Agent protocol for plain hosts, and gains a Cap'n Web transport.
  • A plain Durable Object composed with WebSockets now works with useAgent and AgentClient: the capability sends the identity frame on connect…
  • useAgent({ transport }) and AgentClient({ transport }) pick the wire. "cf-websocket" (default) is the hibernating socket; call()/stub send JSON rpc frames.
Kilo Code
  • #14251 b3ef153 - Add @model, @past-chats, and @worktrees mentions to the Agent Manager New Worktree prompt.
  • #14249 d73b514 - Run worktree-independent slash commands from the Agent Manager New Worktree prompt.
  • #14274 d162846 - Keep the Agent Manager .kilo directory out of git in subdirectory workspaces and avoid creating it in folders that are not git repositories.
E2B
  • 5c0f6eb: Remove SDK-side defaults from API request payloads so the API defaults apply when options are omitted.
  • 5c0f6eb: Remove client-side validation of the fork count argument. The API validates the requested fork count and rejects invalid values.
Mem0
  • Config: Added keyFingerprint to the config schema for install-count deduplication; installs are now counted once per key rather than on every activation
  • Telemetry: Events are no longer delivered twice, and the plugin_version field now reflects the plugin that produced the event
  • Subagent conversation history is readable through the owning parent session
  • Configurable live subagent status and read-only transcript switching from the interactive prompt
  • experimental-harness sessions can track todos, shown as a pinned line under the input and in full on /todo, and keep a per-session scratchpad whose notes are re-stated to the agent after the…
  • [Doctor] Add System Doctor with AI diagnosis, system checks, guided fixes, diagnostic export, and feedback.
  • [Browser] Agents and chat assistants can use the built-in browser, with an animated pointer showing Agent actions.
  • [Office] Add the office-transform skill: agents can extract worksheet ranges, document paragraphs, PDF pages, and slide shapes into new files.
No release notes published.
  • bf7e6e50e7322ace62c9c2a28fa322613d4961da MUL-5241 fix(agent): bound the hermes shutdown so an escaped pipe holder can't wedge a turn
  • 91ad87186437230303d7a6862a16089656b4fdbe MUL-6648: feat(cli): add issue comment update command
  • f9f5e3b81fc428729159212899d1ff8ba0f350be MUL-6734 fix(daemon): resolve Windows set-path overrides via LookPath
  • Fork a session into a new writable copy. A local or imported session can be forked with its complete research history — conversation branches, Notebook records, artifact versions, literature…
  • One product name: Open-Science. The app now presents itself consistently as Open-Science across the interface, CLI, and packaging.
  • Session plans survive context reconstruction. After the agent rebuilds its context, the active Session Plan is recovered with its identity, revision, and pending approvals, and plan tools stay…
  • fix(csp): enforce worker module import policies
  • fix(csp): match wildcard ports on exact hosts
  • fix(csp): parse each policy in response header lists
  • Chat moved to the left sidebar: the left column now has Assets and Chat tabs, and the inspector stays visible on the right while you talk to the agent.
  • Chat polish: a chat opens scrolled to its latest message, attachments sit above the message as outlined chips, and a message with only attachments no longer renders an empty bubble.
  • Canvas edits always reach the source: dragging or editing a prop in the editor now overwrites whatever the JSX held for it, an expression (x={MARGIN}) included.
  • commit: 40162c508ff10b0f5d3398fd4af5db82dd39f319
  • source tag: beta2026-09-18_15-22-31
  • regions: cn global
  • A second agent harness, behind a flag. pi (pi-agent-core) now runs in-process in the sandbox alongside OpenCode, selected per project by the pi_harness feature flag.
  • The pi_harness setting reads correctly in all nine languages.
  • Internal: the sandbox daemon's timing budgets now tolerate a slow machine during a process respawn, so a correct restart is no longer reported as a failure.
  • SEC EDGAR serves US company statements as they stood on the run's date: a period that has ended but has not been filed is not served, and a figure restated later still reads as first reported.
  • A historical run is no longer served a present-day company profile by either fundamentals vendor.
  • Dated tools take the run's date from graph state, so an omitted or later date cannot reach a vendor.
  • Add TypeSafeModel for TypeSafe's Jev
  • Pass xhigh effort through on Bedrock when the model profile supports it
  • Allow gpt-5.6-sol, gpt-5.6-luna, and gpt-5.6-terra on Bedrock Converse
  • Added Remote Control (start a connection via --remote-control startup flag or /remote-control slash command) to create a session-scoped remote connection for following and controlling your active…
  • Changed the default timeout for headless (-p / --prompt) runs from 5 minutes to unlimited so long-running agent turns run until the response completes unless --print-timeout is passed explicitly…
  • Improved headless (-p / --prompt) error reporting when a turn terminates on an agent or model API failure: the CLI now prints a structured AGY_ERROR: {...} JSON line on stderr with canonical…
  • Added BaseAudioResampler.flush() and BaseAudioResampler.reset(), so callers can mark stream boundaries themselves rather than relying on SOXRStreamAudioResampler's inactivity timeout.
  • Scripted eval expectations on function_call accept an eval:: each matched call is judged by name and arguments, under a judge prompt of its own, so a scenario can check what args: cannot match…
  • AWSTranscribeSTTService.Settings gained partial_results_stability, which selects AWS Transcribe's "high", "medium" or "low" interim-result stability. It still defaults to "high".
  • Updated bundled Claude CLI to version 2.1.276
  • Cline starts again after updating. In 0.0.31 the app could fail at launch with "desktop backend exited before publishing its endpoint" and never reach the workspace picker.
  • Tell new users to run opensre to get started instead of opensre setup. (c8ed2de18) — Vaibhav Upreti
  • fix(ci): align README install contract with documented methods (eb04ee1fb) — Vincent Hus
  • Update README.md (6b7d8c015) — Vincent Hus
  • Godot 4.2+
  • Active Ziva account
  • Captain assignment, scenario and tool controls, and improved playground testing. [Enterprise]
  • Guided WhatsApp setup and clearer account health.
  • WhatsApp campaigns and calls for BSUID-only contacts.

Sep 17, 2026

Kortix
  • Choose which account a connector runs as, per call. A connection now carries the accounts you may run it as, and the account is chosen when the connector is called rather than pinned to the whole…
  • An ambiguous connector call is refused, not guessed. When several accounts are reachable and none is named or pinned as default, the call is denied with account_required instead of silently…
  • Choosing a connector account is discoverable in the CLI rather than a dead end.
Claude Agent SDK
  • Updated bundled Claude CLI to version 2.1.275
Claude Code
  • Added the signed-in account to Claude apps gateway sign-in: when the gateway names it, you confirm it before the credential is saved, and /status shows it
  • Added a send-now key (ctrl+enter, or ctrl+x ctrl+s) that interrupts the current turn and sends all queued messages at once; sent and queued messages show in gray until the model receives them
  • Added a startup warning when a configured otelHeadersHelper fails, so sessions that silently export no telemetry are noticed
  • fix(core): align conditional approvals with validated tool arguments
  • fix(core): deliver tool-not-found output on server-managed resume
  • fix(sandbox): keep command paths POSIX on a Windows host
  • More reliable cost history: publish validated Codex reporting windows sooner, preserve prior data during incomplete scans, and correct Claude cost estimates.
  • Clearer usage meters: recover Cursor Enterprise member budgets, Kimi Code API quotas, Copilot seat-credit bars, and OpenCode Go Monthly selection.
  • Smoother everyday use: improve update access, privacy, shared-card dates, and recovery from temporary provider failures.
Cline
  • Cline can now work on a remote machine over SSH, with the app itself staying local.
  • Sub-agents spawned in the same step now run at the same time instead of one after another, so delegating three independent pieces of work takes as long as the slowest one rather than the sum of…
Planning With Files
  • resolve-plan-dir.ps1 and attest-plan.ps1 no longer call [IO.Path]::IsPathFullyQualified, which does not exist on Windows PowerShell 5.1, so a PWF_PLAN_ROOT pin works on 5.1 again.
  • The PowerShell resolver tests paths literally; a project path containing [ or ] resolves its plans under pwsh.
  • A zero-byte .active_plan no longer raises in resolve-plan-dir.ps1, ledger-append.ps1, ledger-summary.ps1 and phase-status.ps1.
Hypit
  • resolution is required by the endpoint while the model's port is optional, so a request stating none is sent at 2K, the resolution the HypiHub Provider also selects.
  • The endpoint takes no adaptive ratio for text-to-video, while frame mode resolves the framing from the uploaded image and reference-to-video defaults to adaptive.
  • H3 returns its video at content.url rather than the ModelArk video_url.
Omnara
  • feat: offload oversized tool results to artifacts
  • feat: Add docs links and add event log view to chat
  • feat: add custom model providers and link to add models from grant dialog
  • Route Desktop to state-machine loop via ACP prompt meta
  • EUrouter as a declarative provider
  • Connect external backend ACP socket to redirect-resolved backend URL
No release notes published.
  • Feed a test from a Test Manager dataset row — pass --dataset-id and --dataset-row and every column in that row becomes a ${col} parameter inside the run.
  • Rows are addressed by ID, never by position — a dataset row is identified by its _id, so reordering or inserting rows upstream doesn't silently change which row a run picks up.
  • Clear errors when the flags don't line up — the grid runner checks --dataset-id / --dataset-row before starting instead of failing mid-run.
  • feat(api)!: reject legacy reads for new organizations
  • feat(ci): deploy AI Gateway automatically to staging
  • feat(api): attribute legacy GET rejections to clients
  • bb76eb653a65152901cfcc769b46d7a7596b93ab feat: add magic link sign-in
  • fbcb496f887690ab42b95bbe62438fb474d01852 fix: keep empty forceLanguage empty so browser language detection works
  • 0db629eead826afd51bc8dce1a71760b13bc38a5 fix: keep resolved RADIUS/push provider id through MFA setup enable step
No release notes published.
  • feat(viewer): restyle session detail sections
  • feat(action): opt-in resolution of the bot's own outdated review threads
  • feat(viewer): integrate a shared SVG icon set
  • MCP over HTTP. genoffice mcp --http <port> serves the same tool set as a Streamable HTTP server for agents on other machines: a container, a sandbox, a shared box.
  • Docs: picture-heavy documents open lazily with media served from the file on disk, so gigabyte files open in seconds; files over 512 MB are refused with a message instead of a crash.
  • Markdown: saving keeps unchanged blocks byte-for-byte and re-serializes edited blocks in the document's own conventions (opt-in unchanged-source round trip
  • fix(test): keep the Codex fake dependency-free — inline its atomic dump write
  • fix(chat): an edited message shows at once, not after the reply
  • fix(vps): several turns per bot at once — only the desktop is exclusive, and only once used
  • #14209 3eaf47b - Remove KiloClaw: the VS Code chat panel and sidebar button, the /kiloclaw TUI command and chat view, the Kilo Chat client and token handling, and the kilo.claw.status and…
  • #14204 d202c80 - Animate the Kilo logo on the welcome screen: hover the mark to see it turn and play the yellow slot-machine animation.
  • #13519 9e61d68 - Show ChatGPT Codex quota alongside other provider usage in the CLI and VS Code.
  • feat(agent-profiles): select secrets available to a profile
  • feat: forward Docker conversation runtime settings
  • feat(automations): select saved agent profiles
  • #3849 34ec5d2 Thanks @wbxl2000! - Add the /desktop slash command (alias /install-desktop) and the kimi install-app subcommand to open the Kimi Code desktop app page in the browser.
  • #3851 faec32c Thanks @Grapedge! - Render mermaid code blocks as diagrams in the terminal; turn it off under /settings → Mermaid diagrams, or set mermaid = "off" in the [markdown] section of tui.toml.
  • #3838 f4e5822 Thanks @sailist! - Fix an occasional crash when a running turn is canceled.
  • Fixed payment errors from non-Zed model providers incorrectly prompting users to upgrade to Zed Pro instead of displaying the original provider error message.
  • Fixed different snippet extensions for the same language cancel each other
  • Add copy button to code blocks
  • fix(composer): clarify failed attachment drops
  • Distinguish pane load failures from empty states
  • 未指定模型的成员现在跟随主会话当前使用的模型,不再回退到硬编码 Opus。
  • Agent 页面上的模型别名(如 opus)会经过供应商映射解析,在第三方供应商下落到该供应商自己的模型槽位。
  • 在 /config 中显式固定的成员模型仍然优先,行为不变。
  • Added automatic subagent guidance for custom agents defined in Markdown: agents that list invoke_subagent in their tools now receive the roster of available subagents and usage instructions in…
  • Improved background task naming so a task keeps the name it was given when sent to the background; completion notifications and task lists now show that name instead of a generic auto-derived one.
  • Fixed stale credentials lingering after the sign-in server definitively rejects them (for example a revoked or invalidated enterprise account).
Pydantic AI
  • GHSA-vmxc-h2x2-jmf3 (moderate): the cloud-metadata and private-IP blocklists could be bypassed with an IPv6 zone identifier on a URL opted into local network access, via…
  • GHSA-fpf4-vwcp-v4hp (moderate): web_fetch processed responses in superlinear time on the event loop, in both the HTML conversion and the charset decode, so a single attacker-chosen page could…
  • GHSA-22h6-qm39-v87j (low): web_fetch_tool's domain lists were compared as written rather than in the form the resolver uses, so a blocked domain could be reached under another spelling.
  • Windows R is dependable again. Conda-managed R starts reliably on Windows, the interpreter is resolved again after environment materialization, verified kernel recovery preserves execution, and…
  • Notebook turns replay correctly. Same-turn inputs and standard-library imports are restored on replay, so a rerun sees exactly the inputs the original run saw.
  • Reviewer corrections keep their context. Linked feedback threads no longer lose correction context between rounds, and automatic review survives starting a new conversation.
  • 修复历史会话被误判为空、从会话列表中消失的问题;纯图片消息和未完整读取的日志不再被误判为空会话。(
  • 修复回退编辑后旧消息重复发送、待处理队列累积的问题,避免恢复回退分支时再次执行历史排队内容。(
  • 优化长对话的流式渲染,修复输出卡顿和尾部异常空白。(
  • chore(deps): bump nodemailer from 9.0.3 to 9.1.1
  • chore(deps): bump sharp and wrangler
  • chore(deps): bump @vitest/mocker and vitest
  • commit: 2451b101e5693f97814206f778db01f125052186 (upstream baseline)
  • built commit: 1a9fa16196f85aded894d8f598411d6ea3685d99
  • source note: GitHub 源码归档对应上游基线;安装包基于 fork 构建提交,两者源码不完全一致。
  • fix(geometry): unify cross-frame coordinates and screenshot conversion | 统一跨 frame 坐标与截图转换
  • refactor(observation): decouple DOM processing from capture orchestration | 解耦 DOM 处理与采集流程
  • refactor(observation): unify cross-frame capture coordination and data ownership | 统一跨 frame 采集协调与数据归属
  • fix(work-items): schedule reminders at exact requested time (64133252a) — Devesh
  • fix(repl): render delegated CLI children at the replay width so tables stay intact (c77154a1d) — Vincent Hus
  • fix(shell): delegate command syntax to host shell (5081e92ba) — Anwesh

Sep 16, 2026

  • Support aliases as connection identifiers
  • Record reasons for deployment creation failures
  • Collect human feedback and pause events in tracing
  • Live voice now runs the call end to end, computer use reports what it sees at every step, and phone callers can interrupt whenever they like.
  • Ask out loud to end or mute the call. During a live voice session you can now say things like "end the call" or "mute for a second" and your assistant just does it, no button hunting required.
  • "Take a look" turns on your camera or screen share. Ask your assistant to look at something mid-call and it switches on screen share or the camera itself, so showing beats describing.
Deep Agents
  • Fixed prompt clipboard word deletion behavior.
  • Clarified unknown model effort in footer.
  • Prevented duplicate clicks in the footer model and effort pickers.
OpenMausBot
  • fix(test): stop asserting a fixed order on concurrent Fish voice requests
  • fix(coordination): name the bot id coordinate_bots could not reach, and point to list_bots
  • fix(computers): let Windows use "This computer" like macOS
  • refreh your extension and all CoS plugins in chatgpt
  • !OpenAI email
  • !OpenAI email
agent-browser
  • Fixed recording cursor and mouse movement timing so cursor rendering stays synchronized with page content during drags and timed mouse moves
  • Added Vercel Labs product and project status badges to the README
Kortix
  • Projects with thousands of sessions no longer slow the app down. The session list now loads one page at a time — 50 sessions, with a Load more control — instead of returning every session on every…
  • Opening, sharing, stopping, restarting or renaming a session now works however old the session is.
  • Date grouping (Today, This week, Older), filters and search work as before, applied to the sessions you have loaded.
Hypit
  • Local rendering prepares, inspects and launches the same explicitly selected browser. Profiles can select an exact browser version, an existing executable or one archive mirror.
  • Workers and capture processes resolve the installation actually being executed. Rendering preserves element animation, waits for visual resources, improves process cleanup and samples Studio…
  • Windows media execution retains the required process environment, OAuth preserves the complete authorization URL, and yt-dlp locates its Python service through its declared package.
  • Add conversation_link chat field rendered into the system prompt
  • Surface MCP structuredContent and drop null optional params
  • Add configurable request timeout for New Relic toolset
  • feat: scope an agent profile to specific MCP servers
  • feat: add GPT-6 Astra model support
  • feat: track authenticated Canvas arrivals
  • fix(selection): respect document and modal inert boundaries
  • fix(dialog): apply modal UA styles and top-layer state
  • fix(selection): preserve collapsed inline whitespace
  • Control agents, panes, workspaces, and worktrees on saved SSH machines with herdr --machine <label-or-id>. Commands use the saved machine's session without needing an open Herdr window.
  • Connect to Windows SSH hosts from Linux, macOS, or Windows. Interactive setup can install or update the complete Windows package after confirmation; background reconnects never install updates.
  • Edit names, filters, and search text at the cursor instead of only at the end. Herdr inputs now support character and word movement, Home/End, deletion, and familiar Ctrl+A/E/K/U/W/Y shortcuts…
  • Full Changelog
  • Cheap model profiles are removed. The second execution mode that let recovery and overrides run on a cheaper model is gone from adapter metadata, agent runtime configuration, task overrides…
  • Agent APIs no longer return plaintext credentials. Every endpoint that serializes an agent — detail reads, the company agent list, and create/update/lifecycle routes — used to echo…
  • Automatic productivity reviews are retired. The detector that turned run counts, comment counts, and elapsed time into management-review tasks is deleted along with its continuation holds —…
  • fix(task-session-manager): recover stopped sessions and refuse silent task_id drops
  • fix(orchestrator-wake): continue lifecycle prompts in the session's current selection
  • fix(task-session-manager): deliver revive prompts with queue semantics
  • Azure OpenAI Responses: Added AzureOpenAIResponses to use the Responses API with Azure OpenAI deployments. See cookbook.
  • Elasticsearch: Added Elasticsearch vector database with vector, keyword and hybrid search. See cookbook.
  • DocumentationMarkdown: Added a transform for Knowledge.sync_pages that converts Mintlify and Fumadocs components into plain Markdown. See cookbook.
  • Queued messages can now be sent into the running turn with Enter. While the agent is working, pressing Enter with an empty composer steers the first message in the queue, so it is picked up at the…
  • The macOS Dock icon now has the standard margin other apps use. The bundled icon and the selectable runtime icons (Classic, Chip, Hologram, Midnight) were edge-to-edge, so Cline rendered visibly…
  • The session sidebar's sort toggle now shows the mode a click switches to, not the one already active — a folder icon while sorted by time, a clock while grouped by project
No release notes published.
  • Changes since langchain==1.4.0
  • release(langchain): 1.4.1 fix(langchain): preserve open MCP object arguments fix(langchain): correct InterruptOnConfig documentation
  • Feature: Elixir alias/import/require/use targets now resolve onto the module's defmodule node across files, so the internal module dependency graph is no longer dropped as dangling.
  • Feature: a Rust self.method() call now resolves to a method defined on the same type in another file (the common split-impl-block layout), pooling methods across every impl of one type and…
  • Feature: a Ruby member call obj.foo on a known-type receiver now resolves to a method foo inherited from a superclass, including across files, using the same conservative promotion as the…
  • Added the on_new_window setting to choose whether new windows show the Launchpad (launchpad, the default) or an empty untitled buffer (empty_tab). (#63522; thanks albertbogusz)
  • Fixed ctrl-tab jumping to random documents when the mouse moved during a quick tab switch. (#52671; thanks OmChillure)
  • Fixed a crash in editor: rotate selections forward and editor: rotate selections backward when using cursors on nonconsecutive lines. (#63937; thanks timvermeulen)
  • fb38eaea363ba8a17818d751950beecd9b1ee5ff feat: add group add/delete actions to group tree page
  • 39a1bba16e8be5cf8169156dc7a04fdf9af73865 feat: support popup mode for Captcha providers
  • 0f629ebed9727b191e259d82a9ed16871abc67fa fix: URL-encode OAuth state on interactive login, consent, prompt, signup and post-logout redirects
  • The PageIndex SDK, local or cloud — vectorless, reasoning-based RAG, end to end.
  • Much faster indexing — the PageIndex Flash engine gets the tree from layout stats: no LLM involved for the structure generation itself, LLMs only write the node summaries, and tree expansion…
  • Flash engine: the local default (mode="standard" keeps the classic LLM pipeline).
Skill Recorder
  • Fix Microsoft enterprise sign-in so it completes SSO first, then opens GitHub Copilot app authorization in the same default browser, without a device code.
  • Make both browser handoffs explicit: after SSO reaches the Microsoft enterprise page, close that tab, return to Skill Recorder and choose Authorize Copilot.
  • Pin the bundled GitHub Copilot CLI to reviewed version 1.0.78, which supports browser OAuth for desktop subprocesses without a terminal. The previously bundled 1.0.71 does not support --web-flow.
  • fix(geometry): unify cross-frame coordinates and screenshot conversion | 统一跨 frame 坐标与截图转换
  • refactor(observation): decouple DOM processing from capture orchestration | 解耦 DOM 处理与采集流程
  • refactor(observation): unify cross-frame capture coordination and data ownership | 统一跨 frame 采集协调与数据归属
  • More reliable usage displays: keep prior measurements visible through temporary connection failures for Codex, Claude OAuth, Cursor, and DeepSeek.
  • More accurate cost estimates: avoid duplicate Claude proxy usage and price OpenCodex records using their recorded provider, keeping unknown costs distinct from zero.
  • Clearer allowance bars: restore Amp Agent and Orb usage and keep Grok weekly pacing visible near reset.
Planning With Files
  • Completed plans no longer emit the routine completion notice through the shared Stop gate or Codex Stop hook.
  • Explicit check-complete commands without the gate flag still report completion. Incomplete-plan notices, gate block decisions, recursion protection, block caps, stall diagnostics and plan…
  • The shell and PowerShell changes are synchronized across maintained canonical mirrors.
Langfuse
  • security: Resolve redirect paths with the URL constructor (v3 backport of
  • fix(ingestion): merge same-timestamp events in arrival order (v3 backport of
  • fix(shared): propagate GCS source-stream errors through pipeline() (v3 backport of
Open Code Review
  • feat(action): add optional pr_number input and workflow_run fallback
  • fix(diff): flag untracked binary files as binary in workspace mode
  • perf(diff): skip oversized untracked files before reading
  • 1648f10ac039e126635e26d4a16647331b2d6f0c: feat: add plan mode (@vadiminshakov)
  • 051fb37ae1d6bd6d6f2f6cbc8a629d5f85aa9bd7: fifix(ui): color the to-do task name while the tool is still running (@taciturnaxolotl)
  • 2019da14717e4e00aeee7664873c6b78f147a3eb: fix(tui): unfreeze spinner clock when the boot busy probe lands (@andreynering)
LoopForge
  • fix: keep executable bits when installing Classic hook scripts
  • fix(desktop): restyle the app icon and regenerate crisp multi-size native icons
  • feat(terminal): make the Windows shell configurable
  • fix(web): refresh live session telemetry
E2B
  • 5b015ad: Removed the hidden e2b template build command (alias bd). It printed a V1 deprecation notice and exited 1; the V1 build system it fronted is gone from the API.
  • Updated dependencies [956e3ab]
  • Updated dependencies [e1fbe86]
  • Share scan settings across the CLI and SDK
  • Accept large saved post-scan prompts
  • Make cost display optional
  • New: MCP support. AI agents can now drive GenOffice over the Model Context Protocol, two ways
  • genoffice mcp — the bundled CLI serves every command as MCP tools over stdio, plus a guided flow for building presentations. Works with Claude Code, Claude Desktop, Cursor and any MCP client.
  • A local MCP server inside the app lets an agent edit the Word document you have open, live. Contributed
  • fix(worktrees): retire the chat tab of a chat with no child when its workspace goes
  • fix(browser): press keys through a US-layout CDP key table instead of a subprocess per keystroke
  • fix(workspaces): re-seed a gate-reported empty workspace after an agent selection
  • Added a /skills reload subcommand to asynchronously reload discovered skills and slash commands without restarting the session or blocking user input.
  • Added an inline argument autocomplete dropdown when typing /model and a live fuzzy search bar inside the interactive /model picker to filter models by name or ID.
  • Improved authentication error messages when OAuth tokens lack required scopes by including explicit instructions to run /logout and /login.
  • commit: 40162c508ff10b0f5d3398fd4af5db82dd39f319
  • source tag: beta2026-09-16_10-58-25
  • regions: cn global
  • the sandbox owner keeps live mutable state and a private snapshot backing
  • the daemon's mount request carries that backing root, so a v0.1.6 sandbox owner
  • the public CLI is unchanged (layerfs --version prints layerfs 0.1.6), and
  • feat(analytics): expand product events for install, tools, Ask User, and workflows (959885a2c) — Vaibhav Upreti
  • chore(deps): bump pyjwt from 2.13.0 to 2.14.0 (3e72293f8) — dependabot[bot]
  • chore(deps): bump types-requests from 2.33.0.20260712 to 2.33.0.20260906 (bdeaf354c) — dependabot[bot]
  • Review skill updates as a real diff. Update previews render through a shared source diff viewer — file headers with change counts, line-number gutter, dashed red deletion and solid green addition…
  • Side chats start as drafts with the right model. New side chats open an empty draft immediately, inherit the main conversation's current model and reasoning effort, and apply your own selection on…
  • A standalone reset utility for Windows. When damaged data survives reinstalling, a guided command-line tool lists resolved data, configuration, profile, and authenticated runtime caches, refuses…
  • feat(components): upgrade Guardrails with optional combined rule and model checks
  • fix: preserve restricted-component diagnostics in streamed errors
  • fix: stop double-prefixing catalog policy CHECK names under naming conventions
  • mempalace rules no longer accepts --agent. Pass --host, --harness and --project, then re-render any installed shared-brain block.
  • get_collection() refuses names MemPalace never reads. Any name other than the configured drawers collection or mempalace_closets raises CollectionNameMismatchError.
  • The MCP event listing returns the newest events first when no cursor is given. mempalace_event_list and palace_coordinate now default to newest-first unless since_event_id or order is passed.
  • Removed the inactive-session collapse (Alt+I): inactive sessions always render in the agents view, and search remains the filter.
  • Changed the default Prime Inference model from GLM 5.2 to GLM 5.3.
  • Changed Agents View session statistics to use muted secondary text while preserving the available details (ENG-6000).

Sep 15, 2026

Omnara
  • feat: Subagents
  • feat: add path-based file transfer tools
  • feat: show model api usage
  • Add Symphony release skill
  • Scrub GitHub and GitLab authentication token aliases
  • Use Bearer authentication for GitLab API requests
Kortix
  • Previews open again when you are signed in. A sandbox preview opened from a session (prod-p<port>-sbx-….p.kortix.com) showed "Sign in to open this preview" even though you were signed in.
  • The release gate quarantines TUN-6 on deployed targets only: Bun's WebSocket client sends no User-Agent, so the edge firewall refuses the tunnel handshake on staging and prod.
  • chore(release): staging VERSION → 0.13.16 [skip ci]
  • snapshot option for system prompts: Added a snapshot field to SystemPromptPreset and a new SystemPromptCustom typed dict.
  • Updated bundled Claude CLI to version 2.1.273
  • fix(core): improve destination validation and connection routing in web fetch utilities
  • fix(core): enforce RFC 9207 issuer identification in MCP OAuth flow
  • chore(release): bump version to 0.60.0-nightly.20260901.g0bd1d4397
Claude Code
  • Added x-claude-code-request-class, x-claude-code-agent-type, x-claude-code-prev-tool-durations, x-claude-code-compaction and x-claude-code-context-compacted request headers for LLM gateways.
  • Added a notification when an MCP server disconnects mid-session and automatic reconnection gives up, pointing at /mcp
  • Added forking a session started with claude --remote-control or /remote-control from the Claude app; the fork runs as a background session on your computer
Strands Agents
  • feat(storage): add bm25 search strategy (495d1540e)
  • feat(sandbox): include partial output in shell timeout errors (bd4020f8d)
  • feat(vended-tools): port Python notebook improvements to TS (6c7988831)
  • Run loop blocking detection. The agent now detects code that blocks the
  • Faster agent startup. The prefork server preloads libraries before
  • Improved agent lifecycle tracing. Traces now cover the full agent
Hypit
  • Each HypiHub request now identifies the Hypit version that sent it, so version-specific service problems can be traced from the request itself.
  • Every HypiHub request carries user-agent: hypit/<version>, read from the installed distribution's own version. This covers API calls, OAuth token refreshes, reference uploads and asset downloads.
  • bash npm install -g @hypit/hypit@0.1.10
  • Title "# Gas City vX.Y.Z"; optional 1-paragraph framing intro. Section order (settled template as of
  • Bullets open with a bolded operative sentence, then dense narrative prose (1-5 sentences);
  • Breaking changes live inside Upgrading Notes as prescriptive imperatives with the exact command
  • A native terminal workbench — Start with nanobot, switch sessions, paste supported clipboard images, complete app and skill references, inspect context and file diffs, queue follow-ups, and use…
  • Browser conversations side by side — Arrange up to four topics in resizable groups, keep sidebar ordering, drag saved sessions into mentions, and use temporary chats when you do not want a…
  • Automations that are easier to manage — Switch between task and calendar views, inspect run replies, and adjust schedules with clearer controls. Editing a job preserves pending executions.
No release notes published.
  • Claude Integration: Access the step-by-step reasoning process of Claude models by ensuring visible thoughts are requested during adaptive thinking tasks. (ba0d542)
  • request visible thoughts for Claude adaptive thinking (ba0d542)
  • Feature: Terraform module calls with a literal local source (./… or ../…) now resolve to a directory-scoped module node, exposing the caller→implementation topology (e.g.
  • Feature: a bare implicit-self call in a Ruby method now resolves up the inheritance chain to a method defined on a superclass, including across files.
  • Feature: an inline code span in a Markdown file (e.g. Widget.render ) now emits a references edge to the symbol it names, but only when exactly one corpus callable matches — generic words and…
  • fix(internals): compare both matrices' column counts in cosineSimilarityMatrix
  • feat(memory): add DakeraMemory for persistent cross-session agent memory
  • fix(examples): migrate basics_colab and agents_colab notebooks from granite3.3 to granite4.1:8b
Kilo Code
  • #14155 f894108 - Show message timestamps on hover in the chat transcript, including the turn's finish time and duration.
  • #14168 a16e5e9 - Show code added with "Add to Context" as a collapsible context card in the prompt input instead of pasting the path and code block as text
  • #12714 23ecd42 - Move marketplace catalog, installation, removal, and installed-item detection into the CLI backend so editor clients use the shared marketplace API.
  • Replace subscribeQueuedMessages({ resourceId, threadId }, listener) with subscribeThreadEvents({ resourceId, threadId }, listener) (listener now receives typed events like queue-count-changed).
  • @mastra/archil: ArchilFilesystem.grep() renamed to diskGrep().
  • Added the Studio Workflow Builder backend. Configure the editor with the new workflowBuilder option to enable a hidden, editor-owned agent that authors persisted workflow definitions
  • Set up sharing in one guided flow — kane-cli context sync setup walks you through sharing over GitHub, an S3-compatible bucket, or a plain folder, or joining a team's existing context.
  • Refusals read like sentences, not stack traces — an unreachable location, a denied account, or an empty location each come back as a named refusal with the remedy, and bucket or Git output is…
  • Setup keeps what you typed — after a refusal your address and credentials survive, a changed bucket re-asks only for keys, and the closing line gives you an address you can paste.
  • The OS drop handler is the desktop's, and only the desktop's
  • Terminate process on window close (orphan concat.exe leak)
  • AppImage, .deb, .rpm and .msi beside the archives
CopilotKit
  • This release repairs several v1 runtime surfaces that were silently broken since v1.50.0 and removes the deprecated useRenderTool shim from React Native.
  • 🔀 Work across GitHub repositories: track multiple PRs in one session, switch between their summaries and diffs, and link or unlink PRs manually.
  • 🎛️ A smoother chat interface: configure workspace, agent, model, effort, and permissions in the redesigned composer.
  • 🪟 Better Claude context management: Claude SDK sessions now show context-window usage and support /compact for native compaction, including with Claude Pro/Max logins.
  • fix(emulation): preserve locale and timezone semantics
  • refactor(emulation): coalesce process environment notifications
  • feat(emulation): comprehensive Emulation support
  • Correct Cursor pace: keep monthly allowances and Grok Bot’s weekly reserve tied to their own reset windows, with restored paid and trial usage.
  • More complete spending totals: retain known spend beside unpriced sources and restore shared costs and charts beside web dashboard account groups.
  • Quieter Claude checks: avoid duplicate session warnings and prevent background usage probes from creating Remote Control sessions.
  • 023c349f5f8de922437dba07449f5f1c95af26d4 MUL-7072 chore(db): drop the reference_only column
  • 8bf525eb2dde0e4e97174bc21ecbc6ff11fc998a MUL-7106 feat(dingtalk): add contextual replies and task lifecycle reactions
  • 7ebe0bf58d99238ccf830c7ed1aa658d4f5807d0 MUL-7213: feat(server): Triage has no executor
  • 标签页:支持拖拽排序、批量关闭、重新打开已关闭的标签;终端可在侧边与底部之间移动而不中断。
  • 文件与输出:文件树单击打开预览、双击固定常驻,同一个文件不会重复开标签;输出卡片整行可点,中文文件名不再点击无反应。
  • 内置浏览器:菜单改为系统原生弹窗,打开时网页保持可见;地址栏支持历史建议,新窗口链接在当前任务内开成后台标签。
  • feat(viewer): add back navigation to improve page traversal UX
  • feat(opencode): add backgroundFile input and report termination signals
  • fix(suggestdiff): preserve case and indentation changes
  • fix(users): stop the column order and visibility keys colliding
  • chore(ai-gateway): Cache rust builds
  • chore: fix test: measure deleted-mask cleanup by physical rows, not patch parts
  • Fix active-plan display and listing when .planning/.active_plan starts with a UTF-8 byte order mark, as written by some Windows editors and PowerShell UTF-8 workflows.
  • The fix is synchronized across all 18 canonical shell helper copies. Regression coverage includes the current-plan display and saved-plan listing with no newline, LF and CRLF endings.
  • @kuei51307-hub for the fix and regression coverage in #243. The contribution is recorded in CONTRIBUTORS.md.
  • Add source-generated GLM-5.3 and GLM-5.3-Flash picker entries for Z.AI Coding Plan, Zhipu AI, and Zhipu AI Coding Plan.
  • Refresh related provider metadata, including GLM-5.2/5.3 highspeed entries where the source exposes them. Picker membership follows the maintained source's tool-calling eligibility.
  • GLM-5.3-Flash exposes image input and a 1M-token context window in generated source metadata.
Cline
  • The app no longer gets stuck on "Desktop backend unavailable" when the backend is slow to start.
  • Confirming text with a Chinese or Japanese IME no longer sends the message. The Enter that commits an in-progress composition also reached the composer, so choosing a candidate fired off a…
  • The model picker refreshes its list every time you open it, instead of staying frozen until you restart the app.
  • npm: bash npm install -g base44@latest
  • Homebrew: bash brew install base44/tap/base44
  • Version: 0.1.15 NPM Tag: latest
  • #3780 486dcd2 Thanks @Grapedge! - Update the terminal UI engine, fixing link colors in wrapped markdown tables and @ file-completion ordering, and adding native clipboard support on Linux X11.
  • #3776 48fad6e Thanks @huangzheng2016! - Fix memory not being released when subagent scopes are disposed.
  • #3777 dc76b0c Thanks @huangzheng2016! - Fix progressively slower rendering on each round of large agent swarm runs.
  • fix(browser): decode Chromium SameSite storage values (135 cookies silently lost per profile)
  • fix(workspaces): complete a worktree create when a post-create step throws
  • Fix landing footer menu placement
  • commit: 1d7aaa2fb4456c1ba96ab7f2125567d7c19bbf69
  • source tag: beta2026-09-15_11-59-37
  • regions: cn global
Caveman
  • Plugin: the Claude Code plugin and Gemini extension manifests declare the release version, so hosts stop showing a commit sha or a years-old 1.0.1.
  • Skill: an explicit reply-language instruction wins over the conversation's dominant language.
  • Stats: /caveman-stats reports only what the transcript shows; the unmeasured savings claims are retired from docs.
  • Added the /copy btw slash subcommand to copy the full text of the active /btw side-question response to the clipboard even when the response card is collapsed or scrolled, along with argument…
  • Improved tool-calling fidelity by translating per-request tool-choice constraints (any, required, and named functions) into native backend function-calling configurations.
  • Fixed the /hooks command and hook inspection utilities omitting hooks bundled inside enabled plugins when listing active hooks.json configurations.
  • feat: support nibrun deployment
  • feat(kuaidi100): add Kuaidi100 provider
  • feat(slack): add OAuth scope selection and real-time search
  • feat(merge): readable conflict review: overview, verdict per conflict, merged code only where combined (2fa06d180) — Yauhen Bichel
  • feat(merge): stream the coding agent's steps while it resolves conflicts (095adcf28) — Yauhen Bichel
  • feat(shell): resolve merge conflicts with a reviewed, approved merge that finishes as OpenSRE Agent (13ae497ec) — Yauhen Bichel
  • Private bookmarks for your reading. Save any passage of conversation text, a text preview, or a PDF text/region selection as a session bookmark with an optional note.
  • Side chats become independent tabs. Several side chats can now run under one session, each as its own named preview tab, while the main conversation stays fully interactive.
  • A redesigned settings center. Seventeen panels regrouped into four purpose-based sections — Intelligence, Connections, Workspace, System — with settings-wide search that deep-links to the right…
  • #780 bf7b08b Thanks @caseyg! - feat: add ibm bob as a provider and coding-agent host
  • #869 9b35620 Thanks @radekfojtik! - fix: add claude opus 5 to the anthropic and vertex model lists
  • #870 f71581b Thanks @pranaypolishetti26! - feat: add kiro coding-agent integration

Sep 14, 2026

CopilotKit
  • Align the package version with CopilotKit 1.71.2. The API and implementation are unchanged from 0.1.0.
  • Added fast mode in Claude Code Remote sessions (cloud and self-hosted runners): the host's fast-mode setting or /fast typed in the session applies where your organization allows it
  • Added mouse support to the /config panel in fullscreen mode: the wheel scrolls the settings list, a click on a setting's value changes it, and the row under the pointer is highlighted
  • Added claude self-hosted-runner --drain-marker-file <path>: when that file exists at a SIGTERM drain, the runner reports its exit to the server as a host drain (telemetry only)
  • Revert model updates for patch release
  • Roll back MCP default version selection update #6d8152a
Deep Agents
  • Update read_file output formatting.
  • Surface DeepSeek V4.1 Flash in the model picker.
  • Surface locally tracked GitHub stacks in agent context.
Chat ON Steroids
  • Supports effort-only picker labels such as Medium, High and Extra High without requiring the old 5.6 prefix.
  • Handles model-version rows with retirement captions, including the new leaving-date message.
  • Recognizes the closed 6 Pro label while validating the actual provider model identity.
  • fix(pptx): speaker notes scale to match slides
  • @stephenswetonic made their first contribution in
  • Disabled subagents for claude code and codex
  • A small patch release with one focused fix for setups where the assistant hands work to child processes.
  • What's fixed Credentials now survive child-process handoffs. Some tools run their work through sanitized child processes, and those processes were coming up without the HTTP credentials they…
  • Restored ACP session model, effort, mode, and reasoning chunk boundaries when loading, resuming, or forking sessions. (@JacobNWolf)
  • Show remote config authentication errors during startup and exit with a failure status.
  • Request summarized adaptive thinking for GitHub Copilot models.
  • Argument errors point to the relevant help command in both text and JSON output. Unknown help topics report the mistake directly.
  • Result history pagination retains the original project and Source filter by keeping the current query options.
  • Unavailable execution logs return a nonzero exit code with guidance on project and Runtime selection. An existing empty log remains a successful read.
  • Remote dashboard sessions no longer expire on refresh bursts (#110061, fixes #55712; salvage #71548 @Doud-FR, #55717 @liuhao1024).
  • Long-lived processes stop leaking duplicate state.db writer handles (#110934, fixes #100896 #103339.
  • hermes update (git installs), or re-run the installer one-liner.
  • feat(ai-gateway): relay authenticated OpenAI Responses requests
  • feat(filters): replace search scope dropdowns with inline suggestions
  • feat(table): use one column popover across all tables
  • 🐛 Resolve variables in Forge option arrays.
  • 🐛 Strengthen preview isolation, linked draft permissions and preview session creation.
  • 🐛 Restrict WhatsApp sessions to internal runtime access.
  • add --resource flag to analyze a single resource (6863057)
  • add ValidatingAdmissionPolicy analyzer (c520d4a)
  • anonymise resource names in event-derived failures (1813838)
  • 6f670995436a0bab99e3fdddd0eebc5c52c39072: fix(home): require path boundary when shortening or expanding ~ (@teddytennant)
  • 066f4444d086aeac877ec0a67eb2a3d56fedde27: openaicompat: bump fantasy to fix regression around reasoning-only turns (@meowgorithm)
  • feat(linear): filter Updated Issue trigger by changed fields @AdamSelene, @claude, @OdaiAhmed99
  • fix(pieces): declare remaining runtime deps for datadog and scrapeless @kishanprmr
  • fix(pieces): hide body input for GET/HEAD requests @kishanprmr
  • #3749 6126472 Thanks @liruifengv! - web: AI session titles are now always on — a title is generated after the first turn and can be regenerated from the rename field, with no experimental flag…
  • #3670 a9efbe0 Thanks @Grapedge! - Delete sessions from the session picker: press Ctrl+X on a session, then y to confirm.
  • #3763 dd6a411 Thanks @liruifengv! - web: add user agreement and privacy policy entries to the Settings → About page.
  • feat(git): validate if PATH git version is safe git
  • fix(rules): route Python stub files to the Python review rules
  • fix(preview): show provider-excluded diffs
  • docs: changelog and blog post for v0.9.2
  • fix(retain): batch a document's chunk embeddings instead of one request per chunk
  • perf(tokenizer): replace tiktoken with quicktok and default to o200k_base
OpenViking
  • 新增会话提交参数 reset_context:清空后续注入的会话上下文,保留 Session ID、原始历史和长期记忆,并阻止后续摘要继续继承重置前的 overview。OpenClaw reset hook 已接入。
  • OpenClaw 根据宿主版本选择 afterTurn 或 commitTurn 捕获消息;持久投递路径在写入失败时向宿主返回失败以便重试。召回改用本轮传入的 prompt,避免用历史消息代替当前问题。
  • OpenClaw 可选择 commitRetentionMode: "turn_budget",按用户轮次和 token 预算保留近期上下文;同时修复相关 pending_tokens 计数。
  • Reliable Codex costs: restore estimates after session rescans, preserve pricing across interrupted scans, and retain saved history during upgrades.
  • Antigravity CLI recovery: restore quota reporting with current agy releases while preserving selected-account boundaries.
  • Clearer provider credits: keep balances beside reset dates, restore explicit credit-pool percentages, and show available Warp add-on credits after monthly credits run out.
  • Upload a local app build straight from your laptop — point kane-cli at the build sitting on your machine and it gets uploaded for you, no separate upload step or dashboard trip.
  • The uploaded build is wired into the run automatically — the resulting app id is forwarded to the test, so the run uses exactly the build you just sent.
  • remote sends emulator and simulator tests to HyperExecute — the same command that runs locally can now execute on remote Android emulators and iOS simulators.
  • chore(task-session-manager): drop generation fence, dead branch, and per-event board sort
  • fix(v2): restore /preset, stop stray orchestrator wakes, and drop beta-era paths
  • fix(task-session-manager): stain only consumed unresolved takes and gate adoption labels
  • Open Interpreter 0.0.43
  • Maintenance release based on upstream Codex rust-v0.154.0. Adds easy choices for Gemini 3.8 Flash and 3.7 Flash, Z.AI ZCode and GLM support, and inherited GPT-6 Astra plus GPT-5.6 Sol, Terra, and…
  • Multi-Agent session selection: the in-session multi-Agent selector adds a lead-Agent marker that clearly separates the lead Agent from members, while simplifying the view in single-Agent setups.
  • Collaboration fix: fixed failures when delegating tasks from non-lead Agents in multi-Agent group chats, now correctly identifying member Agents.
  • Member Agent image rendering: fixed images generated by non-default Agents failing to render on the Web console and desktop client.
  • Godot 4.2+
  • Active Ziva account
  • Speed gains vary with session length and usage. Longer sessions improve the most.
  • Shell failures now give the model clearer recovery guidance.
  • /trace now provides better diagnostics for session-title issues.
  • For the installer path, run CodeWhaleSetup.exe; it installs
  • Extract the archive for your machine: codewhale-windows-x64.zip or
  • Double-click codewhale.bat (not the raw .exe) to launch
  • d60b5ae57429ce987eccf0e616c96b38e8c5423f feat: add configurable password history check
  • e1aa11639006165122fd1dc55edf1140359275ba feat: add organization-level defaultTokenFields for new applications
  • d6f2cb6dddbc5a1f0deb152b9d8e9c5bc5aa4c19 feat: show device, IP and last active time for each session
  • GA 版本正式发布
  • 更新检查默认仅查稳定版,可选包含预发布
  • 登录页与浏览器标题 slogan 更新为「懂你、帮你、陪你成长的智能伙伴」
  • docs(shell): document five missing slash commands and the picker's stay-open menus (9f91c395a) — Awokoya Olawale Davidson
  • feat(observability): add opt-in Langfuse tracing for turns, LLM calls and tools (ecd20b785) — Vincent Hus
Omnara
  • fix: remove api prefix on MCP server hosts and fix well-known path

Releases by week

Every stable release of the week, Monday to Sunday, for the last 26 weeks. Each agent's full history is on its own page.