Initialize Tessl in a repository, and anchor it there. tessl init detects the coding agents present and configures them automatically — Claude Code, Cursor, Codex, Gemini CLI, Antigravity, GitHub Copilot CLI, and GitHub Copilot for VS Code need no further setup; anything else that speaks MCP is configured by hand, with shipped recipes for OpenCode and OpenClaw.
The integration itself is deliberately thin: an MCP server entry (command: tessl, args: ["mcp", "start"]) plus written rules files. That is how one skill reaches seven harnesses without being rewritten — “Write context once; Tessl distributes it everywhere” — and it is the mechanism behind the integrates_with: edges on the tessl page.
tessl project handles the second half: create, link, repair, and list the project, which is Tessl’s stable anchor for a repository. It exists so repository-connected results — evaluation runs above all — stay attached to the right codebase as branches and checkouts come and go. The three nouns are kept deliberately apart: a plugin is content, a workspace is ownership, a project is location.
tessl project repair earns its own subcommand, which is a small tell about how often the anchor breaks in practice (renamed repos, re-clones, forks).
See Also
- tessl-workspace — the ownership boundary a project belongs to.
- tessl-eval — the results a project exists to anchor.
- beads-setup — the store layer’s equivalent: a per-harness install recipe rather than auto-detection.