Back to directory

GitMesh

An open-source multi-agent orchestration runtime and governed MCP server that runs AI agent teams against a repository under a maintainer-defined policy-as-code layer.

GitMesh

GitMesh is a multi-agent orchestration runtime for open-source projects that runs pre-defined agent roles, such as triage, PR review, docs, security, community, onboarding and release, against a repository. Every connected tool, including Claude Code, Copilot, Cursor, Codex and Gemini CLI, is governed by a single policy-as-code layer defined in YAML and compiled to Rego through Open Policy Agent, so no agent merges a pull request, edits CI/CD files or publishes a security advisory without human approval.

The orchestration engine handles atomic task checkout, persistent agent context, heartbeat scheduling and budget enforcement, and synchronizes bidirectionally with GitHub or GitLab issues and pull requests over webhooks. An orchestrator coordinates multiple agents over JSON-RPC so they can work simultaneously without conflicting or duplicating work, and every action is recorded in an immutable, filterable audit log.

A three-view dashboard covers active agents and their budgets, a mobile-first pending-approvals queue, and the audit log. Adoption needs one YAML file and one CI step; it is written in TypeScript and licensed under Apache-2.0.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.