AI agent releases: Jun 15 – Jun 21, 2026
154 stable releases from 81 tracked agents and frameworks in the week of Jun 15, 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.
Jun 21, 2026
- MiniMax: recover detailed token-plan windows from the remains API when the coding-plan page only exposes coarse usage. Thanks @Yuxin-Qiao!
- Cost history: remove the redundant tooltip from submenu-backed Cost rows. Thanks @Zihao-Qi!
- Menu refresh: keep the menu open and show in-place progress when Refresh is clicked. Thanks @elijahfriedman!
- Add configurable .pr_agent.toml branch selection via CLI/env with GitHub fallback by @[copilot-swe-agent[bot]]( in #2411
- test: strengthen focused unit coverage for core behavior
- fix (A2A): fetch public diff for PR URLs and report failed state to Langfuse
- Search relevance updates that return more helpful results for common queries.
- More robust ingestion and background sync so your added content is indexed reliably.
- Lower query latency — searches and retrievals complete faster in typical workloads.
- Fix packaged skill frontmatter.
- Add regression coverage for skill metadata.
- Package version bumped to 0.1.3.
- For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
- Extract codewhale-windows-x64.zip
- Run install.bat (copies to %USERPROFILE%\bin)
- Richer Telegram delivery: Telegram now sends rich HTML, preserves rich markdown and sticker paths, renders progress drafts and command output more faithfully, normalizes HTML tables safely, and…
- More dependable agent recovery: retries, terminal outcomes, usage after compaction, session history repair, and reply reconciliation now keep more interrupted or partial turns moving toward a…
- A stronger Codex integration: Codex gains automatic plugin approvals, GPT-5.3 Spark OAuth routing, remote-node exec as a dynamic tool, and more reliable app-server teardown and terminal outcomes.
- Feat: Cocoindex show full fledged inspection tool for debugging
- LanceDB improvements
- docs(lancedb): fix callout syntax, document vector/FTS indexes, tighten prose
- Sandbox package - Added @agent-browser/sandbox with shared, Eve, and Vercel Sandbox helpers, example projects, and docs for running agent-browser in hosted sandbox environments
- Updated sandbox release flow so the new package stays version-synced with the CLI release and publishes from the correct workspace path
- Reflowed documentation prose across the README, docs site, examples, and skills so Markdown and MDX wrap naturally in editors and renderers
- Honor configured agent step limits by forcing a final text response instead of failing mid-run.
- Fix Devstral model detection when provider IDs use different casing. (@Robin1987China)
- Pass configured custom headers to Copilot model requests.
Jun 20, 2026
- Updated bundled Claude CLI to version 2.1.185
- Chat-template thinking compatibility - OpenAI-compatible custom providers can map Pi thinking levels into chat_template_kwargs, enabling vLLM/Hugging Face chat-template models such as DeepSeek to…
- GLM-5.2 provider improvements - GLM-5.2 now has corrected Fireworks OpenAI-compatible routing and OpenRouter xhigh thinking support, improving /model behavior and high-effort reasoning for GLM-5.2…
- Added inherited configurable chat-template thinking support for OpenAI-compatible providers that use chat_template_kwargs, such as DeepSeek models behind vLLM.
- Thank you @0xhis for fixing AppImage glibc mismatch crashes by ensuring the bundled ld-linux is always replaced from the host system.
- Full Changelog
- 11662010521e5211d825fcc21c4fdf8c72fda08e: fix(bangmode): activate bang mode when ! is preceded by whitespace (@meowgorithm)
- 7e4bd6a0b44ce3c32f0e36d48ffcd77c1955d641: fix(bangmode): engage bang mode when pasting text starting with ! (@meowgorithm)
- f342edf052ef1ca35fb202c235a976b9ef6f45f9: fix(bangmode): include bang commands in history (@meowgorithm)
- wren-core: support composite (multi-column) primary keys (d33917f)
- wren: GenBI app build & deploy — semantic layer → shareable web app (67257bf)
- ensure UTF-8 encoding for YAML file operations on Windows (322bd79)
- fix release packaging
- update overlay
- Merge pull request #55 from palmier-io/codex/lightweight-sentry-telemetry
No release notes published.
- Widgets: add single-window and combined burn-down charts for Codex and Claude session/weekly limits. Thanks @jamesjlopez!
- AWS Bedrock: show optional rolling 14-day Claude token and request totals from CloudWatch. Thanks @zyaiire!
- Codex: optionally show both session-window and weekly percentages in the compact menu bar label. Thanks @thepraggyverse!
Jun 19, 2026
- feat(assemblyai): add universal-3-5-pro (now default) and Voice Focus streaming params
- feat: add Gemini 3.1 flash TTS support, implement streaming response …
- (google): fix formatting
- Plugins are now a first-class marketplace. Browse, install, and update plugins from inside the app, alongside the ones the vellum team ships.
- Privacy policy and consent updates. We have updated the privacy policy and the Share Analytics / Share Diagnostics control.
- Workflows - When your assistant needs to apply the same operation across several things (researching products from different brands, comparing multiple options, summarizing a batch of docs), you…
- We tweaked the font colors to make it more legible;
- A long-running command will now show immediately and stream output as the command runs;
- Session name generation is fixed for sessions started with a command.
- Hermes reaches iMessage — Photon Spectrum, no Mac relay required — There's now an iMessage platform plugin built on Photon's managed line pool.
- Raft — Hermes joins the Raft agent network as a gateway channel — A new bundled Raft platform adapter lets Hermes connect to Raft as an external agent through a wake-channel bridge.
- A substantially more capable desktop app — v0.16.0 shipped the desktop app; v0.17.0 deepened it across dozens of PRs.
- None called out in this changelog.
- Random bump
- Improved state signal restoration so long threads can resume without scanning every message.
- Commands in /help are now listed alphabetically in both the CLI and ACP
- /teleport is now always available and shows an explicit error when its prerequisites aren't met, instead of being hidden
- Firecrawl Research Index — Added a specialized index for agentic AI/ML research: search across 3M+ arXiv papers and the GitHub code behind them (issues, merged PRs, and READMEs, refreshed daily)…
- Keyless access for core endpoints — Use /scrape, /search, /interact, and /parse without an API key from official MCP, CLI, and SDK clients.
- Automatic PII redaction — Added a redactPII option that strips personal and sensitive data like names, emails, phone numbers, addresses, and secrets out of scraped content before it's returned.
- add Anthropic API support (1c4e77a)
- allow Azure OpenAI API version override (1b7d1f0)
- deps: update k8s.io/utils digest to a95e086 (8da4716)
- Free, no key — tushare, okx, yfinance, akshare, baostock, tencent, mootdx, ccxt, futu, local, plus four new direct-API additions: Eastmoney, Sina, Stooq, and a direct-HTTP Yahoo client.
- Optional key-gated US providers — Finnhub, Alpha Vantage, Tiingo, FMP; absent keys are skipped in the fallback chain, never crash it.
- ~98 non-merge commits since v0.1.9
- Godot 4.2+
- Active Ziva account
- Selective provider base entry points - SDK users can pair @earendil-works/pi-ai/base and @earendil-works/pi-agent-core/base with explicit provider registration to keep bundled applications from…
- Mistral prompt caching - Mistral sessions now use provider-side prompt caching with session affinity and cached-token usage/cost accounting. See API Keys and Environment Variables.
- Post-compaction token estimates - Compact results and compaction events now include estimated post-compaction token counts so clients can show the approximate context reduction.
- Merge pull request #38 from palmier-io/htin1/filler-words
- Rescale clip duration when set_clip_properties changes speed
- Add ripple_delete_ranges and get_transcript MCP tools
- feat: add pre-approval tool input guardrails
- feat: add SDK-only custom data for tool outputs
- fix: enforce the documented strict JSON-compatible contract for
- Updated bundled Claude CLI to version 2.1.183
- 🤖 More agent harnesses & SDKs
- run agents on Google Antigravity (Gemini 3 Pro, also Claude/GPT-OSS)
- Cursor via a first-party Python SDK and a native cursor-agent (ACP) harness with cost tracking
- bb desktop 0.0.20
- ci: cache Flutter pub packages in the Mobile job (ceb5eff53)
- Delete RESEARCH directory (a1d5570d3)
- feat(cli): support ephemeral channels via --ttl on create/update (4f671a255)
- Cascading config edits: delete_with_cascade and rename_with_cascade now ripple through agents, providers, channels, model providers, and skill bundles instead of leaving dangling references.
- Discord skill-driven slash commands: dynamic slash commands generated from installed skills, with typed options, chunked interaction followups, edit/delete/bulk-delete archive sync, and…
- Operator Console redesign: the gateway web UI ships a schema-driven config editor, multi-agent chat, read-only skills browser, and a long list of ported fixes.
- Added a complete German (de-DE) interface for both the desktop and mobile apps, so German-speaking users can now use AionUi fully in their own language.
- Brazilian Portuguese (pt-BR) is now selectable in the settings and login language pickers — the translation existed but couldn't actually be chosen before.
- Fixed blank PDF previews on Windows — opening a PDF from the workspace file tree now renders inline instead of showing an empty panel.
- #2178 c49a3fc Thanks @seanmcguire12! - add support for WebMCP
- #2217 147e310 Thanks @monadoid! - Add Azure OpenAI Microsoft Entra ID model auth support.
- #2231 cf3603d Thanks @miguelg719! - Add claude-fable-5 support: native structured outputs via the @ai-sdk/anthropic bump, adaptive thinking (including the new "xhigh" effort) on the agent path…
Jun 18, 2026
- bb desktop 0.0.19
- Add Relink to reconnect offline media
- Show 'Media Offline' instead of black for unloadable clips
- Restrict MCP HTTP server to localhost with origin validation
- Bump version to 1.0.43 (2a4a20c05)
- SKY-11180 Fix Copilot directory synthesized code binding (d02ead94f)
- Make free credits a one-time grant (ae05ec184)
- Update: programbench uses v6 images
- Skills Store - Browse, install, and manage agent skills from a dedicated in-app store, so you can give your agents new capabilities without hand-wiring files.
- Self-hostable sandbox execution - Cloud/sandboxed agent execution matures: a self-hostable Kubernetes sandbox provider plugin lands alongside server-side Kubernetes execution integration and…
- Per-company multi-tenant isolation - A major security foundation for shared and cloud deployments: each company now gets its own JWT signing keys, cloud tenants are strictly company-scoped…
- release(langgraph): 1.2.6
- fix: nested subgraph inherits parent checkpoint_ns (regression in 1.2.3)
- fix: cancel running subgraphs on v3 stream abort [closes #8029]
- Preserve Registered Model Params on Reconstruction: Reuse the live
- cookbook: update AG-UI examples to OpenAIResponses and gpt-5.4
- fix: preserve registered model params when reconstructing agents
- Changes since langchain==1.3.9
- release(langchain): 1.3.10 chore: bump cryptography from 46.0.7 to 48.0.1 in /libs/langchain_v1 chore: bump aiohttp from 3.14.0 to 3.14.1 in /libs/langchain_v1 fix(langchain): switch summary…
- Call external APIs as first-class test steps — flows can now include execute_api steps that dispatch a named API call, store the response, and pass it forward to later steps in the same run.
- Child flows inherit API context — when a flow spawns a child, the child has access to the parent's API registry and writes its response back so the parent can read it.
- API variables resolve by dot-path — output values from an API step can be referenced with dot-path syntax in subsequent conditions and actions, including inside if_else branches that follow an API…
- add create_http_options to ContextCacheConfig for cache creation timeout (ffc9677), closes #4703
- add GCS first party toolset to ADK integrations (fb19e1a)
- Add log_level option for adk run CLI (1ac69a9)
- fix(excel-html): chart overlays match Excel position and size
- Automatic theme mode - /settings can choose separate light and dark themes and follow terminal color-scheme changes. See Selecting a Theme.
- Self-only updates by default - pi update now updates pi only, with pi update --all for updating pi and packages together. See Install and Manage.
- Extension API helpers - extensions can use CONFIG_DIR_NAME for project config paths and import edit diff helpers for edit-style diffs. See ctx.cwd and SDK Exports.
- feat: build CLI during pull request
- fix(helm): build chart dependencies before lint
- test(e2e): remove python gpu smoke test
- Fixed .gitignore handling edge cases: ignored .gitignore rules stay active, trailing-slash ignore-control patterns no longer leak the file, and descendants of a directory literally named…
- Fixed duplicate relative paths when packing multiple roots.
- Updated root dependencies, including major bumps to v15.
- refactor(auth): introduce plugin-based authentication architecture
- fix(sdk/go): expose search date and level filters
- fix(ragfs): fix root directory acquire encryption account_id exceptio…
- cognify_status now works in API mode — the MCP can resolve dataset IDs remotely and return pipeline status.
- CogneeClient can query dataset/pipeline status via the REST endpoint /api/v1/datasets/status when running in API mode.
- Vector store (LanceDB) retrieve() now returns an empty list for empty id inputs instead of erroring.
- fix(app-center): stabilize factory selects
- feat(desktop): add tuttiExternal.logs.write for workspace app web.log
- feat(desktop): polish external import and settings action UI
- fix(scripts): correct prerelease arguments expansion and bump version to 1.1.12.post1
- fix(memory): rename ChromaDB probe collection to 'probe-test'
- fix(proactive): prevent cache pollution of load_agent_config()
- docs: changelog and blog post for v0.8.2
- docs: add Memory Defense to the 0.8.2 blog post
- blog: Cursor persistent memory (editor + CLI in one post)
- PyPI: bash pip install crawl4ai==0.9.0
- Docker: bash docker pull unclecode/crawl4ai:0.9.0 docker pull unclecode/crawl4ai:latest
- Note: Docker images are being built and will be available shortly. Check the Docker Release workflow for build status.
- agent-framework-core: Add AgentLoopMiddleware for re-running agents in a loop
- agent-framework-core: Integrate tool approval into the harness agent
- agent-framework-core: Add tool approval middleware
- Fixed the Agent exiting prematurely under the default exit_conditions=["text"]. The agent now only stops when the last message is an assistant message with non-empty text…
- Scheduled task management: View, edit, enable/disable, and delete any scheduled task directly in the console. The task list is sorted by enabled status first, then by next run time.
- Knowledge base categories and document management: The knowledge base can now be organized by category, with documents under each category managed in the UI. Thanks @yangziyu-hhh
- Multiple custom model providers: Configure multiple OpenAI-compatible providers and switch the active one with a single click, fully compatible with existing configuration. Thanks @kirs-hi
- refactor: add search pipeline hooks for context rendering
- Dev 20260615 v2.0.20
- For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
- Extract codewhale-windows-x64.zip
- Run install.bat (copies to %USERPROFILE%\bin)
- Q&A Page: Optimized voice input logic to append transcribed text instead of overwriting existing input content;
- Q&A Page: Added a clear input content function after users type content in the input box;
- Q&A Page: Users can normally send new questions after switching to Session 2 while Session 1 is generating replies;
- fix(desktop): stop dimming deferred message lists (718b596e5)
- Smooth channel loading: single-surface timeline state machine (d47572209)
- feat: surface base + persona system prompts in observer feed (5004758fb)
Jun 17, 2026
- Thank you @Huyuochi for adding a Kimi provider “Get API key” affiliate link to streamline setup.
- Thank you @Ron Liu for enabling Telegram inbound voice messages with transcription support.
- Thank you @Sam for tightening numeric grounding rules in agent prompts (with regression coverage).
- Session timelines load much faster and avoid flicker or scroll jumps.
- OpenAI-compatible providers now accept MCP tool schemas that previously failed validation. (@jquense)
- Cloudflare AI Gateway now receives the configured API key correctly. (@keefetang)
- Resilient DB Component Loading: Isolate each agent/team load so one bad component is skipped instead of dropping all, fix model provider round-trip on deserialization, register TuningEngines, and…
- Structural Registry Tool Dedup: Dedupe re-instantiated toolkits by (type, name, function set) so they collapse instead of accumulating duplicate registry entries.
- ParallelTools GA API: Migrate to the parallel-web >= 1.0 GA surface.
- Add v0.2.22 to appcast
- Memory tools are now native OpenCode tools registered via the @opencode-ai/plugin tool() helper and backed by the mem0ai SDK directly.
- Skills load via the config hook (skills.paths) instead of being copied into the project's .opencode/ directory on startup.
- Trimmed to 9 focused skills (context-loader, dream, forget, status, search, scope, pin, remember, tour).
- feat(oci): skip startup full scan on unchanged logic
- feat: LanceDB vector and FTS indexing support
- fix(lancedb): escape single quotes in delete filter predicates
- Agent: Added auto-compaction and /compact to Zed Agent.
- Agent: Improved agent skills management by moving it into the settings UI.
- Agent: Improved how we communicate problems relative to skills, in both the agent panel and the settings UI.
- fix(server): retry sandbox delete phase conflicts
- feat(server): support TLS certificate hot-reload
- feat(providers): add DeepInfra as a built-in inference provider
- [BREAKING] @onecli-sh/sdk 0.5.0 → 2.2.1 — requires a OneCLI server with the /v1 API. Older servers 404 every SDK call.
- New agent provider: Codex (OpenAI) — run /add-codex. Full runtime via codex app-server (planning, MCP tools, server-side history, resume). Trunk ships the seams and the skill.
- Setup can now select, install, and authenticate a non-default agent provider. A provider registry feeds the setup picker, an installer pulls the provider's payload from its branch, a vault auth…
- 👌 Configure WhatsApp forwarded events [c5516cd]
- 👌 Configure WhatsApp webhook forwarding URL [3db24c4]
- 🐛 Fix landing page Discord URL [b4a7aab]
- Configure — create and edit assistants, attach skills, manage MCP servers and LLM providers, and change app/UI settings, all by asking.
- Diagnose — when something isn't working, the Butler can inspect your conversations, read logs, and check the health of your providers, scheduled tasks, teams, and MCP connections to tell you what…
- Remote access — ask it to put your local WebUI online and it walks you through a one-click public link (via an automatic Cloudflare tunnel) — no networking knowledge needed, no jargon.
- npm: bash npm install -g base44@latest
- Homebrew: bash brew install base44/tap/base44
- Version: 0.0.56 NPM Tag: latest
- Reverted a the unread counts for mentions, participating and folders due to performance related issues.
- Model provider: Allows users to modify the model type of existing model configurations. #16029
- Chat channel: Enables users to deploy RAGFlow assistants as chatbots within external messaging platforms like Discord and Feishu. #15850
- Observability: Groups multi-turn chat traces by session within Langfuse to facilitate conversation debugging and replay. #15679
- #1767 f03dee6 Thanks @threepointone! - Reduce Think Durable Object SQLite reads during normal wakes and text-only turns.
- #1722 9f8e14b Thanks @mattzcarey! - Fix two MCP client OAuth bugs found by the new conformance suite, and add MCP conformance testing.
- MCPClientConnection now finishes OAuth on the transport that received the 401. A fresh transport loses the resource metadata URL from the WWW-Authenticate header, so token exchange fell back to…
- init auto-installs the curated smithers skill into the coding agents it detects on your machine, so there is no more manual mkdir or curl.
- It seeds an editable hello workflow (.smithers/workflows/hello.tsx with its prompt in .smithers/prompts/hello.mdx) as the smallest end-to-end example to run and tweak.
- The home page and the For Humans guide were rewritten around this one-command flow.
- Models Page Overhaul: Provider Aggregation, Unified Card UI & Layout Redesign
- Simple Mode: Add simple mode with flat nav and sorted session list by updated time
- Wide Mode Toggle: Add wide mode toggle for expanded chat layout
- macOS 签名发布链路进入验证阶段:正式 release workflow 会预检 macOS 证书、Apple ID、app-specific password 和 Team ID;secrets 完整时构建签名/公证的 DMG/ZIP,并额外跑 Gatekeeper package-smoke。手动触发 workflow 时 draft 默认开启,适合先生成草稿 release…
- Provider/runtime 上下文声明更准确:provider 设置支持 1M context 声明和自定义 context window,Anthropic providers 默认带上 Tool Search 能力,新会话会优先使用 provider 的主模型,减少模型能力和 UI 状态不一致。
- 流式请求更稳:首 token 等待会尊重请求超时,整体 stream 会限制 free-trickling provider 的最长持续时间;中途遇到可恢复的 transient tool-call 错误会重试,而不是直接让本轮失败。
- feat: instrument the sign-up/sign-in funnel and move product analytics to PostHog @moaaz-ae
- feat(output-schema): friendly labels for opaque keys and array-rooted outputs @AhmadTash
- fix(worker): raise isolate sandbox open-files cap to 4096 @abuaboud
- Python: Add function_choice_behavior support to Azure AI and OpenAI Assistant agents
- Fix MessagePack
- .Net: fix: prevent duplicate "null" in JSON Schema type arrays for nullable parameters
- test(openclaw): cover install contracts and runtime docs
- Feature/wiki link
- feat(openclaw): add TOS release installer flow
- Opt-in ACP last message snippets for sessions
- Unified OTLP logging schema for cross-tool detection
- Canonical thinking modes across providers
- Added on_user_turn_message_added event handler on LLMUserAggregator, with a new UserTurnMessageAddedMessage arg type.
- Added RealtimeServiceMetadataFrame, broadcast at pipeline start by realtime LLM services (OpenAI Realtime, Azure Realtime, Inworld, Grok/xAI Realtime, Gemini Live, AWS Nova Sonic, Ultravox).
- Added to our examples "locally-driven-turns" variants for
Jun 16, 2026
- add Copilot BYOK provider wrapper utilities and CLI support (e67ee2a)
- add dashboard agent usage stats (6d3f39f)
- Add support for Mistral Vibe CLI (0932b8b)
- Fixed HTTP dispatcher configuration to preserve a caller's deliberate fetch override instead of reinstalling the undici global fetch over it.
- Fixed inherited OpenCode Go DeepSeek V4 thinking-off requests to send the provider's thinking: { type: "disabled" } compatibility parameter.
- [x] Now supports Intel, AMD, and NVIDIA GPUs for a 92% smaller binary and 15% faster processing times.
- If you already have the NVIDIA GPU binary installed, you can safely delete it if you want. It will still work and is backwards compatible.
- [x] Support for Developer API for transcription on audio (POST: /v1/transcription/transcribe)
- A new Mac app. The Electron-based desktop app replaces the legacy Swift build and keeps feature parity with the web UI and iOS.
- Download the new DMG from vellum.ai/downloads
- Login security improvements. All clients got a security pass: more secure login and token management across the board.
- Full Changelog
- Conversation unread counts, sidebar badges, ordering, and filter support.
- New Help Center layout, per-locale settings, category icons, and editor improvements.
- WhatsApp/Twilio calling updates, voice messages, mute controls, and inbox call settings.
- Publish Docker Hub image attestations and document digest verification by @[copilot-swe-agent[bot]]( in #2433
- feat: add Gemini 3.1 stable and 3.5 model identifiers to supported model registry by @[copilot-swe-agent[bot]]( in #2432
- feat(config): add --extra\_config\_url to merge external .pr\_agent.toml
- Support Lottie in inspect_media
- Add Lottie motion-graphics support
- Fix audio volume reset between clips
- feat(desktop): refine thread-unread badge to two-token form (de24b90aa)
- fix(buzz): prevent reconnect storms from reaped ephemeral channels (fd2553726)
- fix(buzz-acp): trim oversized observer frames to fit instead of dropping (5a651632d)
- MCP server - Added agent-browser mcp, a stdio Model Context Protocol server with typed tools, paginated discovery, protocol negotiation, and startup tool profiles.
- Plugin system - Added out-of-process plugin support over the agent-browser.plugin.v1 stdio protocol, with plugin add/list/show/run, manifest discovery, npm and GitHub refs, credential providers…
- Added context footprint eval coverage for CLI skills, MCP core, and MCP full-profile surfaces, plus MCP parity tests to keep tool behavior aligned with the CLI
- gateway-client exports createGatewayCollection, gatewayCollectionDefs, and the run-node helpers (flattenGatewayRunNode, snapshotToGatewayRunNode, reconcileSnapshotNodes), all backed by the gateway…
- gateway-react adds the SyncProvider context, the GatewayCollections registry, and the hooks useSyncQuery, useSyncMutation, and useSyncSubscription, plus the typed shortcuts useGatewayQuery…
- The removed exports have no drop-in replacement. Rebuild custom-UI data access on the collection hooks.
- Richer channel delivery: Telegram and WhatsApp are less brittle: Telegram renders structured text with tables, lists, expandable blockquotes, preserved intentional line breaks, and CLI-backed…
- More reliable agent runs: account-scoped DM sends, generated media completions, auto-reply message-tool final replies, reset archive fallback reads, restart shutdown aborts, yielded subagent…
- Safer model routing: new GLM-5.2 and Claude Haiku 4.5 catalog support arrives with normalized provider IDs, managed SecretRef auth, bounded model browsing, and safer OpenAI/Anthropic tool-schema…
No release notes published.
- Hotfix for a frontmatter regression in v3.1.2.
- The description refresh shipped in v3.1.2 added a colon ("...AI coding agents: keeps..."), and the English SKILL.md carry the description field unquoted.
- Fixed: the description is now wrapped in double quotes across the canonical file and the seven English IDE variants (.codebuddy, .codex, .cursor, .factory, .hermes, .mastracode, .opencode).
- Path-traversal hardening for caller-supplied identifiers. sender_id now carries the same path-safety guard as app_id / project_id: a character whitelist plus rejection of the . / .. tokens.
- Defense-in-depth write containment. MarkdownWriter now rejects any write target that resolves outside the configured memory root before reading, creating parent directories, or writing files.
- Add a multimodal usage guide and correct the multimodal error semantics after end-to-end verification.
- viking://agent/... 仍可读旧数据,但不再是新写入目标。
- 新的 memory / resource / session / skill 数据应进入 viking://user/...。
- agent_id 只作为 legacy 过渡配置使用,会映射到请求级 actor_peer_id。
- #1757 92a5ba1 Thanks @threepointone! - Bump the partyserver dependency to ^0.5.8, which base64-encodes the
- #1754 151d457 Thanks @threepointone! - Pin accepted WebSocket connections to binaryType = "arraybuffer". On Worker
- #1754 151d457 Thanks @threepointone! - Add Browser Run Live View support to the browser tools. The cdp connector
- For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
- Extract codewhale-windows-x64.zip
- Run install.bat (copies to %USERPROFILE%\bin)
- ACP workspace trust gated behind the workspace-trust client capability
- /resume now opens much faster
- Startup update-failure message is now gentler and no longer alarms when an update can't be applied
- Code Interpreter runtime paths have moved from /opt/opensandbox to /opt/code-interpreter.
- Added line-based file reads through GET /files/download with offset and limit query parameters, enabling code-interpreter clients to read large text files by line range.
- Added directory listing support through the new execd directory API. Directory listing supports depth control, reports file, directory, symlink, and other entry types, and does not recursively…
- Bot sees the message but session never starts → no reply ever arrives
- Windows event log / stderr shows GBK-encoded 命令行太长。 (command line too long)
- ❌ 错误: session process is not running may appear in user-visible cards
- ci(webui): add format and build checks
- feat(omni): streaming audio + live captions for DashScope/OpenAI omni
- feat(agent-team): support custom subagent templates in agent service
Jun 15, 2026
- Companion v0.1.3: Prepare and checksum update for companion v0.1.3 release.
- macOS x64 Companion Asset: Added target to companion release workflow (cross-compiled on macOSrunner).
- Installer Preservation: Fixed installer to preserve build and plan agents during install.
- Only send grounding_metadata for 3.1 live at the end of each turn (847a259)
- streaming: Ensure final partial=False frame is always yielded (6e59c61)
- Support parallel-web >= 1.0 GA API in ParallelBackend: migrate web_search/web_extract to the top-level client and pin the floor to >=1.0
- fix: update ParallelBackend for parallel-web >= 1.0 GA API
- chore: release 2.6.16
- Add ask_user_question tool so the agent can pause mid-turn and ask 1–4 multiple-choice questions in the CLI
- Show an inline question widget in the chat (tabbed options, multi-select, free-text "Other", optional notes)
- Degrade gracefully on non-CLI channels (WhatsApp/gateway/subagents) when no interactive user is available
- Stop controls are now slot-scoped: pausing one teammate no longer cancels the whole team, and each pane's send box is gated independently.
- Team chats now prefer the assistant's identity (avatar/name) when one is set, and the team workspace display correctly prefers the persisted team workspace, falling back to the leader's only when…
- Fixed assistant cron and Guid flows: cron tasks now show the right assistant avatar/name, selected skills are kept on Aionrs conversations, and assistant metadata is preserved through team creation.
- ci(stale): add stale issue and PR workflow
- fix(e2e): correct return type of _stub_with_token
- Added local plugin v1 support with plugin.link/list/unlink/enable/disable, manifest-declared actions, event hooks, managed plugin panes, link handlers, command logs, keybinding integration, and…
- Added herdr plugin install <owner>/<repo>[/subdir...], plugin uninstall, source metadata in plugin.list, offline registry fallback, and a human-readable default plugin list with --json for scripts.
- Added herdr plugin config-dir <id> and automatic plugin config/state directory creation so plugin setup docs can point users at a stable config path.
- Fuzzy search in slash command autocomplete
- Syntax-highlighted, line-numbered, theme-aware diffs for file edits
- /resume now scopes the session picker to the current folder
- 5 new first-class agents: Devin CLI (beta.1), Google Antigravity agy (beta.5),
- Major platform expansion: QQ Bot inline keyboards & file send/receive (OneBot),
- New core commands: /timer (one-shot delayed task), /cancel (interrupt
- Chrome no longer hangs on startup — the CDP launch now has a bounded retry loop and a configurable timeout, so a stuck browser process fails fast instead of blocking your run indefinitely.
- Retry logic handles edge cases correctly — previously, certain hung or degenerate branching states could cause retries to stall or misbehave; these are now resolved cleanly.
- {{var}} placeholders in start URLs are passed through as-is — the CLI no longer tries to resolve or rewrite template variables in start URLs before the run begins, so your parameterized URLs reach…
- [ROB-321] Restore regression tag on ES multi-cluster evals + wire ES secrets in…
- Propagate runtime-built llm_instructions through MultiInstanceToolset (fixes Confluence eval 210)
- Add per-request tracing experiment routing and span token metrics to server
- Execd install path flattened from /opt/opensandbox/bin/ to /opt/opensandbox/ — the emptyDir mount in the K8s provider and the Docker bootstrap launcher now use a single /opt/opensandbox directory…
- Credential Vault — new credential injection system that lets sandboxes consume secrets (API keys, OAuth tokens, HTTP headers, query parameters) without exposing raw values to user code.
- Conflicting endpoint modes now rejected — requests that combine use_server_proxy=true with expires (asking for both a server proxy path and a signed gateway route) now return…
- Credential injection: Per-sandbox secrets are attached to outbound requests at the proxy layer via EgressRule.inject — user code inside the sandbox never handles raw credentials.
- Domain filtering: Policy-driven allow/deny lists gate which destinations a sandbox may reach, evaluated first-match-wins against the L7 request.
- Access auditing: Structured JSON logs of every egress request with optional body redaction via a redactor Lua module, enabling downstream compliance review.
- [codex] Add reusable external PR review gate
- [codex] Use configured RD members for external PR gate
- [codex] preserve agent display prompts and fix electron install
- Cursor IDE plugin (.cursor-plugin/). Auto-registers the mempalace-mcp server, five slash commands (/mempalace-help, /mempalace-init, /mempalace-mine, /mempalace-search, /mempalace-status), and a…
- Cursor IDE hooks (stop / preCompact / sessionStart). Background auto-save every N agent turns, synchronous transcript mining before Cursor compacts, and session-start memory recall scoped to the…
- First-class Antigravity (Google) IDE support. New .antigravity-plugin/ package and idempotent installer that registers MemPalace as an Antigravity plugin — MCP server, skill, and two lifecycle…
- Automatic first-run theme selection - pi detects the terminal background on first run and defaults to the dark or light theme. See Selecting a Theme.
- Standalone binary integrity checksums - GitHub release assets now include SHA256SUMS files for verifying standalone binary downloads. See Quickstart Install.
- Added SHA256SUMS integrity files to standalone binary GitHub release assets.
- Full Changelog
- remove upper cap for scale transform
- Add inspect_timeline agent tool
- Add clip-aware AI Edit submenu to timeline clip context menu
No releases match.
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.
- Nov 16, 2026431 releases · 120 agents · this week
- Nov 9, 2026717 releases · 152 agents
- Nov 2, 2026632 releases · 157 agents
- Oct 26, 2026628 releases · 145 agents
- Oct 19, 2026614 releases · 144 agents
- Oct 12, 2026545 releases · 150 agents
- Oct 5, 2026495 releases · 127 agents
- Sep 28, 2026429 releases · 121 agents
- Sep 21, 2026426 releases · 126 agents
- Sep 14, 2026299 releases · 107 agents
- Sep 7, 2026245 releases · 101 agents
- Aug 31, 2026268 releases · 99 agents
- Aug 24, 2026286 releases · 108 agents
- Aug 17, 2026234 releases · 95 agents
- Aug 10, 2026250 releases · 95 agents
- Aug 3, 2026230 releases · 89 agents
- Jul 27, 2026201 releases · 79 agents
- Jul 20, 2026199 releases · 82 agents
- Jul 13, 2026222 releases · 83 agents
- Jul 6, 2026198 releases · 81 agents
- Jun 29, 2026137 releases · 71 agents
- Jun 22, 2026157 releases · 72 agents
- Jun 15, 2026154 releases · 81 agents
- Jun 8, 2026158 releases · 72 agents
- Jun 1, 2026130 releases · 59 agents
- May 25, 2026123 releases · 67 agents