Release v0.9.2
Hey there! We have been busy making Autohand better. Here is what is new since v0.9.1:
New Stuff
- Adding support for account sync
- Expand skill discovery for shared agent locations
- Expand agent import coverage for OpenCode, Kimi, and Gemini MCP
- Add idle logout opt-out for long-running agents
- New feature Introducing --squad option
- Add explicit bare startup mode
- Add configurable status line controls
- Accept provider keys in config set
- Expand reflection parser scenario coverage
- Add GLM-5.2 and GLM-5.1 to Z.ai provider
- Add custom OpenAI-compatible provider configuration
- Add experimental session fork and clone commands
- Add guided goal authoring command
- Add goal writer queue sequencing
Bug Fixes
We squashed 17 bugs:
- Guard session recovery notes before conversation bootstrap
- Handle Ollama responses without message wrappers
- Restore skill mention discovery in the composer
- Fixing regression on cursor use for ink rendering
- Keep composer cursor state aligned during mid-line editing
- Fix startup composer cursor placement
- Prevent sync from corrupting rotated auth tokens
- Handle invalid config set usage explicitly
- Handle startup config parse failures explicitly
- Recreate session directories before writes
- Classify LLM Gateway 400 errors by body
- Restore interactive startup login menu
- Handle reflection as response metadata
- Keep debug notices out of the Ink transcript
- Keep the Ink composer mounted between turns
- Fix custom provider setup verification and display
- Handle noisy provider config and process reports
Documentation
- Document and enforce explicit bare runtime boundaries
- Adding documentation for the new /statusline options
- Update README community and localized docs links
Updates
- Improving and getting ready for Paid plans
- Adding tests coverage
- Render the composer as an open input surface
- Make queued prompts grouped and editable
- Place status line fields in the composer help line
- Rename feature switch command to experiments
- Ignore terminal EPIPE shutdown errors
- Improving some tests and new features
- Adding /exit as well as /quit
- Start authentication without the fragile startup menu
- Stabilize Ink composer footer repainting
- Reflect successful turns into durable memory
- Defer automatic feedback while Ink has queued prompts
- Complete hook event coverage
- Dealing with corrupted history sessions
Under the Hood
Some housekeeping and reliability improvements:
- Align Ink renderer usage with Ink 7 exports
- Disable process auto-reporting in test subprocesses
- Use Ink cursor placement for the composer
- Tighten reflection parsing and runtime input contracts
- Verify composer stability during agent turns
- Verify agent-turn composer stability without loopback
- Linking reference config in their user own language
- Gate handoff session behind experiment
- Gate turn memory diagnostics behind debug mode
- Derive alpha releases from stable tags
- Harden dependency and release automation
Full comparison: v0.9.1...v0.9.2
Get it
Quickest way:
curl -fsSL https://autohand.ai/install.sh | shVia npm or bun:
npm install -g autohand-cliOr grab a binary below for your platform.
| Platform | Architecture | Binary |
|---|---|---|
| macOS | Apple Silicon | autohand-macos-arm64 |
| macOS | Intel | autohand-macos-x64 |
| Linux | x64 | autohand-linux-x64 |
| Linux | ARM64 | autohand-linux-arm64 |
| Windows | x64 | autohand-windows-x64.exe |