# MetaGO Agent Harness

MetaGO Agent Harness is an MCP server for Claude Code and Codex that delivers 55 tools and 8 prompts for hallucination prevention, KMWI memory handling, and runtime skill generation.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 6 repo stars
- Author: metago-ai
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/metago-ai/metagolifeform/tree/HEAD/packages/mcp-server
- Updated: 2026-11-07
- Security check: automated, 2026-11-17 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/metago-agent-harness

## What it does

MetaGO Agent Harness links MCP clients to agent management features over standard IO. It exposes 55 tools and 8 prompts. In Engine V2, three tools execute actual TypeScript routines: `metago_memory_manage` handles a four-tier memory pipeline that persists to JSON, `metago_meta_evolve` drives a five-phase adaptation loop, and `metago_meta_create` writes generated `SKILL.md` files to the filesystem.

The remaining tools output structured prompt guidance. They cover delivery gates that reject incomplete work, multi-level critique against cognitive fallacies, provenance tagging, and source verification. One dedicated tool, `metago_report_event`, relays metrics to the Studio dashboard. Running the server requires Node.js with `npx`.

## Requirements
- Needs: Node.js (commands run via npx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add metago -- npx -y @metago-ai/mcp-server
```

### Codex

**Platform CLI**
```bash
codex mcp add mcp-server -- npx -y @metago-ai/mcp-server
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "mcp-server": {
            "command": "npx",
            "args": [
                "-y",
                "@metago-ai/mcp-server"
            ]
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "mcp-server": {
            "type": "stdio",
            "command": "npx",
            "args": [
                "-y",
                "@metago-ai/mcp-server"
            ]
        }
    }
}
```

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add mcp-server npx -y @metago-ai/mcp-server
```

## When to use it
- Prevent hallucinated APIs and placeholders in agent output
- Store structured knowledge across sessions in JSON files
- Generate new SKILL.md definition files directly to disk
- Validate task completion using multi-layer delivery checklists

## Example prompts
- Run metago_critique on the proposed architecture draft
- Save the extracted conventions using metago_memory_manage
- Execute metago_delivery_gate before closing this ticket

## Similar
- [claude-code-harness](https://affpapa.org/en/nevibecode/plugins/claude-code-harness.md) — plugin · ★ 1596 repo stars
- [swarmclaw](https://affpapa.org/en/nevibecode/skills/swarmclawai-swarmclaw.md) — skill · ★ 710 repo stars
- [pydantic-ai-harness](https://affpapa.org/en/nevibecode/skills/pydantic-ai-harness.md) — skill · 3300 installs
- [meta-harness-optimization](https://affpapa.org/en/nevibecode/skills/meta-harness-optimization.md) — skill · 420 installs
- [bun-runtime](https://affpapa.org/en/nevibecode/skills/secondsky-bun-runtime.md) — skill · 435 installs
- [deno](https://affpapa.org/en/nevibecode/skills/deno.md) — skill · 371 installs

From the same set: [metago-ai/metagolifeform](https://affpapa.org/en/nevibecode/by/metago-ai/metagolifeform.md) — 2 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
