Set up metaharness in Claude Desktop

communitystdiounknown

🛠️ The meta-harness for AI agents — scaffold your own focused, branded agent harness with its own npx CLI, MCP server, memory, learning loop, and witness-signed releases. Works with Claude Code, Codex, pi.dev, Hermes, OpenClaw, and RVM (hardware-isolated sandbox).

01Configuration

{
  "mcpServers": {
    "metaharness": {
      "command": "npx",
      "args": [
        "-y",
        "@metaharness/claude-code"
      ]
    }
  }
}

Add this to claude_desktop_config.json. Generated from the captured install method (npx); the mcpServers shape is shared across Claude Desktop, Cursor, Windsurf, Cline, and VS Code.

02Steps

Make sure Claude Desktop is installed and up to date.
Open claude_desktop_config.json and add the block above (merge into any existing mcpServers).
Provide any required API keys/credentials as environment variables.
Restart Claude Desktop and confirm metaharness’s tools appear.

03Other runtimes

Cursor Windsurf Cline Continue Goose VS Code OpenAI Agents SDK LangGraph CrewAI

04Provenance

config_sourcegenerated from captured install method

last_checked2026-07-06 07:50Z

sourcesGitHub repo search [p4]

Next step

Want agents that act within guardrails? Apex is the live governed-agent product — paced, capped, and fully-logged actions with approvals before anything runs.

Explore Apex →

See also: server page · is it safe? · alternatives