doyaken-cli
Most AI coding tools generate code and hope for the best. Doyaken takes a different approach: it runs an 8-phase pipeline that expands your prompt, triages tasks, plans the implementation, writes the code, tests it, documents it, reviews it, and verifies everything passes — retrying with error context when something breaks.
npm install -g doyaken
doyaken run "add user authentication with JWT"
One command. Working code.
The pipeline
Every prompt flows through eight phases with automatic verification gates between them:
- Expand — turns a short prompt into a full specification
- Triage — breaks the spec into ordered, dependency-aware tasks
- Plan — generates implementation plans per task
- Implement — writes the code using your chosen AI agent
- Test — runs the test suite and captures failures
- Docs — updates documentation to match changes
- Review — code review against quality standards
- Verify — build, lint, test, and format checks; failures loop back with context
The verify gate is where most agents fall down. Doyaken treats it as a first-class concern — if the build breaks or tests fail, the agent gets the error output and tries again, up to a configurable retry limit.
Works with any agent
Doyaken isn't tied to a single AI provider. It works with Claude Code, Cursor, OpenAI Codex, Google Gemini, GitHub Copilot, and OpenCode. Switch agents per project or per task — the pipeline stays the same.
Configuration sync generates and maintains config files for all supported agents from a single source, so your project conventions stay consistent regardless of which agent runs the work.
Built-in skills and prompts
The CLI ships with 40+ reusable skills (security audits, code reviews, CI fixes, vendor-specific tasks) and 25+ prompt templates covering code quality, planning, and development methodologies. Skills can be composed into custom workflows, and Claude Code users get auto-generated slash commands from every skill and prompt.
Self-healing
Production agent workflows need resilience. Doyaken handles model fallback when a provider is down, crash recovery to resume interrupted pipelines, and automatic rate-limit backoff. The goal is unattended operation — start a task, walk away, come back to working code.
MCP integration
Connects to GitHub, Linear, Slack, and Jira through the Model Context Protocol, so the agent can read issues, post updates, and close tickets as part of the pipeline — not just write code in isolation.