cedr

cedr is the command line tool behind cedr.dev, made by Sheernox Technology Group. It picks which AI coding agent, model and effort should take each kind of work, merges reviews from several vendors into a single pass-or-block decision, and pins the skills and MCP servers agents load. The site describes it as in public beta, with Route and Council live and Guard in preview; the README notes commands and file formats may change before 1.0.

What It Is

cedr is a developer tool for people and for the agents working beside them. The CLI works with Claude Code, Codex, Cursor and Gemini, and any agent that speaks MCP can call it directly through cedr mcp. A companion app at app.cedr.dev adds history, sharing and gates on top of the free CLI.

How Route and Council work

Route reads a routes.toml file and picks the agent, model and effort for each kind of work. Equally suited work goes to the plan with the most room left, and plans near their limits are skipped. After the work is done you record a verdict (accepted, reworked, redone or rejected), and cedr route stats shows which model performs best per kind of work.

Council starts one reviewer per vendor, merges their reports, groups findings that point at the same place and decides the gate. Any P0 or P1 finding blocks, and too few reports is never a silent pass.

Guard and pull request checks

Guard lints an agent setup, pins every skill and MCP server in .cedr/guard.lock, and fails CI when a server's tools, descriptions or network hosts change after pinning. A GitHub Action runs cedr lint and cedr verify and shows findings as pull request annotations.

Data handling

The README states cedr never reads agents' credentials and never calls a vendor API for you; plan usage comes from the agents' own status line and CLI. Picks, verdicts and stats stay local unless you configure sync with a cedr.dev API key. cedr push sends names, versions, hashes and findings, not file contents.

Roadmap

The site lists Sync, Meter, Bench and Index as coming soon.