# MetaGO Agent Harness

> MetaGO Agent Harness is an open-source MCP server for ai models and llm tooling. MCP harness with 55 tools for decision locks, memory, self-evolution and provenance. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** macOS, Windows, Linux, CLI
- **Website:** https://metago.life
- **Repository:** https://github.com/metago-ai/metagolifeform/tree/HEAD/packages/mcp-server
- **MCP Registry name:** io.github.metago-ai/metago
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-16

## About

MetaGO Agent Harness is the runtime package of MetaGO, a Chinese project that frames itself as an operating layer for intelligence: criteria written before execution, runtime constraints, and recomputable evidence. The package exposes 55 tools and 8 prompts to any MCP client, covering thinking tools such as planning, quality critique, provenance labelling and value or ethics checks, core skills in governance, safeguards and traceability, and an event reporter for a Studio dashboard.

Three tools run real code rather than returning guidance: metago_memory_manage drives a four-layer KMWI memory persisted to JSON, metago_meta_evolve runs a five-stage evolution loop, and metago_meta_create generates new SKILL.md files from stored patterns. The README says the other 50 tools return structured guidance that the agent follows within its decision lock, and a delivery gate tool blocks declaring a task complete if any verification layer fails.

The vendor is explicit about limits: it publishes 98 release invariants and 6,789 automated tests, says about 89 percent of capabilities are built, and states that adapters are template-level validated. It runs locally over stdio via npx and does not train models.

## Features

- **Decision lock:** Governance tools that bound what an agent may do within stated criteria.
- **Hard-driven memory:** KMWI four-layer memory that persists to JSON.
- **Meta-evolution loop:** A five-stage evolution cycle that records results to memory.
- **Skill generation:** Writes new SKILL.md files from stored patterns.
- **Delivery gate:** A three-layer checklist where any failure blocks completion.
- **Event reporting:** Reports events to a Studio dashboard for metrics.

## Good for

- Add an approval and verification gate to a coding agent
- Give an agent persistent layered memory
- Check a task for completeness before declaring it done
- Label provenance of generated claims
- Experiment with agent self-improvement loops

## Install

### Claude Code (Terminal)

```sh
claude mcp add metago -- npx -y @metago-ai/mcp-server
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "metago": {
      "command": "npx",
      "args": [
        "-y",
        "@metago-ai/mcp-server"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "metago": {
      "command": "npx",
      "args": [
        "-y",
        "@metago-ai/mcp-server"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "metago": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@metago-ai/mcp-server"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "metago": {
      "command": "npx",
      "args": [
        "-y",
        "@metago-ai/mcp-server"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "metago": {
      "command": "npx",
      "args": [
        "-y",
        "@metago-ai/mcp-server"
      ]
    }
  }
}
```

## Tools

- `metago_memory_manage`: Manages the four-layer KMWI memory: add, query, promote, decay detection and health scoring.
- `metago_meta_evolve`: Runs the five-stage evolution loop with time budgets and coupling-score thresholds.
- `metago_meta_create`: Generates new SKILL.md files from internal memory patterns.
- `metago_report_event`: Reports events to the Studio dashboard.
- `metago_delivery_gate`: Three-layer verification checklist where any failure blocks task complete.

## FAQ

### How many tools are there?

55 tools and 8 prompts.

### Which tools run real code?

metago_memory_manage, metago_meta_evolve and metago_meta_create.

### How do I run it?

npx -y @metago-ai/mcp-server over stdio.

### Does it make the model smarter?

The vendor says no; it makes a model harder to misuse and easier to inspect.

### Is MetaGO Agent Harness free?

Yes. MetaGO Agent Harness is open source, so you can use and self-host it for free.

### How do I add MetaGO Agent Harness to Claude Code?

Run this command in your terminal: claude mcp add metago -- npx -y @metago-ai/mcp-server

### What are the best alternatives to MetaGO Agent Harness?

The closest alternatives to MetaGO Agent Harness are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/metago-agent-harness
