Thanks for your interest! AgentBoost aims to stay small, practical, and easy to adopt. Contributions that keep it that way are very welcome.
git clone <your-fork-url>
cd agentboost
npm install
npm run build # compile TS + copy static assets to dist/
npm test # run the test suite (node --test)
npm run dev -- status # run the CLI from source via tsxsrc/core/ Planning, memory, verification, safety, audit, sub-agents + facade
src/mcp/ MCP server (primary integration for agents)
src/gui/ Local dashboard: http server + static SPA (no build step)
src/cli/ Command line tool
adapters/ Config templates for Claude Code / Cursor / OpenCode
examples/ Runnable demo project
docs/ Operation guide (bilingual) + screenshots
test/ Node test-runner tests
- Keep it lean. Minimal dependencies; prefer the standard library.
- Match the style. Read nearby code and follow its conventions and comment density.
- Add tests for new core logic (
test/*.test.ts), and runnpm testbefore opening a PR. - Update docs (
README.md/docs/guide.md) when you change user-facing behavior. - One core, three faces. New capabilities belong in
src/core/and should be reachable from the MCP server, CLI, and (where it makes sense) the GUI.
- Use clear, present-tense commit messages.
- Keep PRs focused; describe what changed, why, and how you tested it.
Open an issue with steps to reproduce, expected vs. actual behavior, and your
Node.js version (node -v).