Claw Orchestrator

Claw Orchestrator gives developers one interface for multiple coding agents and checks their results independently. You define acceptance criteria, and the runtime executes the checks and keeps the evidence on disk.

• Independent verification Checks can cover commands, HTTP responses, screenshots, file assertions and code changes; editing tests used by an acceptance contract refutes the run.
• Persistent sessions Keep agent context across requests and hand a conversation to another engine or model in the same workspace.
• Multiple engines Work with Claude Code, Codex, Antigravity, Grok Build, OpenCode or custom coding CLIs, and ask several engines the same question in parallel.
• Recoverable workflows Checkpointed workflows support resuming, retrying, cancelling and steering work, while a persistent ledger tracks costs across engines and the runtime enforces spending caps.

Claw Orchestrator is open source under the MIT license. MCP connects its tools to compatible hosts such as Claude Desktop, Cursor and Continue. Agent Client Protocol lets editors including Zed and JetBrains use it as their coding agent, and an OpenClaw plugin exposes its tools to OpenClaw agents. A run's acceptance result can be verified, refuted or unverified.