Skip to content
Agents tracked: 258 Downloads (7d): 219M up 6.1% GitHub stars: 5.5M VS Code installs: 148M Releases (7d): 293 Agent status: 2 with issues Updated Oct 7, 2026
AgentGid

Strix changelog: what's new each month

Every stable Strix release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 3 months covered; the current month updates daily.

October 2026 · so far

1 release: 1.7.0

October brought improvements to model handling, HTTP evidence tracking, and configuration flexibility. The release focused on more accurate model matching, better audit trails for security findings, and enhanced compatibility with various LLM providers.

Highlights

  • HTTP exchange evidence can now be linked in reporting for better audit trails.
  • Proxy-validated findings now require HTTP exchange IDs for enhanced traceability.
  • Model frontier checks now match model names accurately without confusing provider routes.
  • New configuration option to force responses versus chat completions API modes.

New

  • STRIXAPITYPE configuration option to control response format modes.
  • Vercel AI Gateway provider guide documentation.

Improved

  • Web search now sends only the agent's query to Exa, reducing unnecessary data.
  • Model compatibility improved with required litellm update to handle gpt-6-astra maxtokens parameter.

Fixed

  • Frontier model checks now match model names only, not provider routes.
  • TUI sidecar hook remains importable with hatchling 1.32.1.

September 2026

3 releases: 1.6.0 → 1.6.2

September brought expanded security coverage, improved reporting capabilities, and enhanced cloud integration for Strix. The month focused on adding new security skills, refining the agent experience, and streamlining configuration handling.

Highlights

  • New security skills added for OWASP LLM Top 10 2026, Azure/Entra, and argument injection detection.
  • Dependency reports now require contextual CVSS scoring and usage evidence for better risk assessment.
  • Caller-provided files can now be placed directly into the sandbox workspace for easier workflow integration.
  • Web search capabilities expanded with Exa provider alongside existing Perplexity integration.
  • Multiple model recommendations updated including GLM-5.3, Fable 5.1, and Gemini 3.7 Flash.

New

  • OWASP LLM Top 10 2026 skill coverage.
  • Azure and Entra security skill.
  • Argument injection security skill.
  • Exa web search provider option.
  • Contextual CVSS breakdown display on dependency reports.
  • Error beacon classification by phase and exception class in telemetry.

Improved

  • Caller-provided files now automatically placed into sandbox workspace via extrafiles and --workspace-file.
  • Resume tokens handled gracefully to improve session continuity.
  • MCP call errors no longer quarantine entire connection.
  • Environment and model checks run on the no-target start screen.
  • Configuration merging improved with environment variables integrated into cli-config.json.
  • Documentation trimmed and reorganized with new cloud CLI and viewer reference pages.
  • Telemetry simplified by dropping per-load skill-loaded beacons in favor of anonymous events.
  • Parent agents no longer wait on finished non-interactive children.

Fixed

  • LiteLLM cost model resolution fixed.
  • Pre-commit mypy hook and test suite now pass on fresh checkout.
  • Cloud top-up instructions now printed on 402 response with guidance for oversized requests.
  • Viewer now correctly shows stopped run status.
  • CLI import warm-up sequencing fixed and orphan purge removed.

August 2026

4 releases: 1.5.0 → 1.5.3

August brought stability improvements and refinements across the CLI, viewer tools, and container handling. The month focused on fixing collisions, resource cleanup, and edge cases in agent execution.

Highlights

  • Fixed viewer tool call collisions that could occur across multiple agents.
  • Added support for custom request headers on OpenAI-compatible endpoints via environment variable.
  • Improved session and resource management in container environments, including browser session reclamation.
  • Fixed several critical agent execution issues including interrupt recovery and tool argument handling.
  • Enhanced UI navigation with clickable mount prompts and better scrollbar visibility.

New

  • Custom request headers for OpenAI-compatible endpoints via LLMEXTRAHEADERS.
  • Docker sandbox containers now labeled with the run ID for cleaner teardown.

Improved

  • Viewer session cookies are now scoped to the bound port.
  • OpenRouter streamed usage cost tracking for accurate billing.
  • Agents can now wait on messages they have already sent.
  • Todo plans no longer fail on priority or duplicate issues.
  • Mount prompt in the UI is now clickable and skips the mount instead of abandoning the scan.
  • Browser environment handling improved to prevent profile permission issues.
  • Empty string arguments are properly coerced to empty containers.
  • SQLite connection management optimized to use per-operation instead of per-thread connections.

Fixed

  • View label spacing alignment in final panel.
  • Recoverable guardrail blocks and crash notification decoupling.
  • TUI scrollbar visibility and report navigation issues.
  • Runner now resumes after user interrupt instead of failing.
  • Browser session abandonment in container environments.
  • OpenRouter app attribution now properly sent on requests.

Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. Strix: pricing, features and alternatives · All changelogs