b26f53c873
* fix(tui): the footer's permission chip outranks the mode word Founder-reported: the "full access" posture chip was missing from the footer in operate mode. The merged footer paints the posture lockup left-to-right and drops any chip that would run past the left block's edge, so the *last* chip was always the first casualty — and the order was [mode, permission]. Operate mode made it reproducible rather than occasional: `operate` is the longest of the three mode words, so it consumed the cells `full access` needed and the footer silently stopped reporting that the session could write anywhere. A preference word must not outbid a safety fact. When only one chip fits, permission keeps the cells; painting order stays mode-then-permission so the lockup does not shift when both fit. Regression test sweeps widths 40..=120 and asserts no width keeps the mode word while shedding permission. Without the fix it fails at width 69 — an ordinary split-pane width, not a degenerate one. tui footer_: 33 passed; 0 failed (existing posture and golden tests included, no golden drift). cargo fmt --check clean. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LmeqaZAesoHjT8N9PR7S2c * docs(agents): code first, then tests Founder rule, 2026-09-01. The implementation is written and proven to run first; tests are added afterward to cover what was actually built. No TDD here, and this overrides any skill or default that mandates it (superpowers `test-driven-development` in particular). An existing test that only encodes old behavior is evidence, not a veto. Placed beside the existing selective-evidence policy it extends, and scoped so it does not cancel the "Claiming a test passed" rule: a regression test written after a fix still has to be shown failing without it. That rule is what makes an after-the-fact test worth keeping, and it is the only part of test-first discipline that was ever earning its cost. No conflicting TDD mandate exists in the scoped crate guidance (checked crates/*/AGENTS.md and docs/). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LmeqaZAesoHjT8N9PR7S2c --------- Co-authored-by: CodeWhale Bot <bot@codewhale.net> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Codewhale
Codewhale is an open source coding agent for your terminal, built in Rust and improved in public with the people who use it.
简体中文 · 日本語 · Tiếng Việt · Bahasa Indonesia · 한국어 · Español · Português · Русский · Українська · Français · Deutsch · 繁體中文 · हिन्दी · Türkçe · Italiano · Polski · العربية · Català
Install
npm install -g codewhale
codewhale
The first run helps you connect a provider or stay offline. Codewhale also supports Cargo, Docker, Nix, Scoop, prebuilt archives, Android/Termux, and a CNB mirror. See the installation guide.
Tab completion is one command per shell — codewhale completion bash|zsh|fish|powershell|elvish.
See shell completions.
Use
Talk to Codewhale the same way you would talk to a teammate:
Fix the failing tests and explain what changed.
Or run a task without opening the TUI:
codewhale exec "fix the failing tests and explain what changed"
Codewhale can read your repository, edit files, run commands, inspect results, and keep working toward a goal. You decide how much access it has.
Why Codewhale
- Use the model you want. Connect hosted providers or local models through
Ollama, vLLM, or SGLang. Switch provider and model with
/model. - Stay in control. Plan is read-only. Ask, Auto-Review, and Full Access make
approval behavior visible.
/undoreverts the last turn and/restorereturns the workspace to an earlier snapshot. - Keep long work organized. Save sessions, set a durable
/goal, review workflows before they run, and coordinate agents without turning their internal instructions into your transcript. - Extend the agent you already have. Connect MCP servers and skills, configure hooks, and keep agent roles as readable files in your project or personal settings.
Run /help in the TUI for commands and keyboard shortcuts.
Safety
Codewhale runs on your machine with the access you grant it. Approval modes and repository rules limit what the agent may do; optional OS sandboxing adds a stronger execution boundary where supported. Unknown model prices stay unknown instead of being reported as free.
Read authorization order for the exact policy stack and configuration for local settings.
Documentation
- Providers and local models
- Agent teams
- MCP, hooks, and configuration
- Local web client
- All documentation
Join the community
Codewhale gets better when people use it, report what feels wrong, and help fix it. If a provider is missing, a workflow is awkward, or the terminal UI gets in your way, open an issue. If you know how to improve it, open a pull request. First contributions are welcome, and contributors keep credit for the work that lands.
Join the Discord, or add Hunter on WeChat
(hunterbown) and ask to join the Whale Brothers group.
Project history
Codewhale began as deepseek-tui and still preserves that configuration and
session compatibility. It is now provider-neutral and independently maintained;
it is not affiliated with any model provider.
Thanks to every contributor and to the open source communities that helped the project grow. See the contributor record.
License
MIT. Portions adapted from other open-source projects are recorded in third-party notices.