# Unified AI System MCP Gateway

> Unified AI System MCP Gateway is an open-source MCP server for developer tools. Self-hosted MCP gateway for Codex, Cursor and Cline with provider-free prompt enhancement. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://happy520ai.github.io/unified-ai-system/
- **Repository:** https://github.com/happy520ai/unified-ai-system/tree/HEAD/packages/mcp-server
- **MCP Registry name:** io.github.happy520ai/unified-ai-system
- **Categories:** developer tools, AI models and LLM tooling
- **Listing updated:** 2026-11-17

## About

Unified AI System sits between your coding agent and execution as a single governed boundary. Its MCP server starts an isolated local gateway automatically, pins it to a deterministic fake provider, and removes the gateway when the session ends. The README is explicit that it does not enable or authorise real provider calls, and the project calls itself a hardened public preview whose verified path uses the fake provider only.

A local deterministic engine can take a rough request and add execution requirements, output constraints, clarification policy and completion criteria while preserving the original text, without calling any model. Read-only tools report gateway health, readiness, knowledge and workforce state and governed agents, while workflow_run executes a three-step local workflow and writes one tenant-partitioned artifact without overwriting existing paths. No tool creates, executes or revokes agents, or approves policies.

It runs from a published container image or from source over stdio or local streamable HTTP, with an optional bearer token and loopback binding by default. The source build exposes sixteen tools.

## Features

- **Provider-free prompt enhancement:** Structures a rough request locally and keeps the original visible.
- **Fake-provider default:** Chat is checked to fail closed unless the real provider flag is false.
- **Governance inspection:** Lists and describes governed agents visible to the authenticated tenant.
- **Local workflows:** Runs a three-step workflow that publishes a tenant-partitioned artifact.
- **Knowledge search:** Searches the local knowledge base by keyword.
- **Stdio or HTTP:** Run over stdio or a loopback streamable HTTP endpoint with optional token.

## Good for

- Try a governed MCP gateway in Codex, Cursor or Cline without any provider credentials
- Turn vague requests into explicit task specs before sending them to a model
- Inspect agent governance and workflow state from an editor
- Evaluate gateway safety behaviour in a local sandbox

## Install

### Claude Code (Terminal)

```sh
claude mcp add unified-ai-system -- docker run -i --rm ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "unified-ai-system": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "unified-ai-system": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "unified-ai-system": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "unified-ai-system": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "unified-ai-system": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0"
      ]
    }
  }
}
```

## Tools

- `gateway_health`: Inspect gateway health and provider safety state
- `gateway_readiness`: Inspect first-run and chat readiness
- `agent_governance_status`: Inspect Agent Governance status
- `agent_governance_list`: List governed agents visible to the tenant
- `agent_governance_describe`: Describe one agent in the tenant
- `gateway_prompt_enhance`: Structure a request locally without a provider call
- `gateway_prompt_enhance_llm`: Enhance prompts via an optional LLM backend
- `gateway_chat`: Send one fake-provider-only chat request
- `knowledge_readiness`: Inspect local knowledge infrastructure
- `knowledge_retrieve`: Search the local knowledge base by keyword
- `workflow_health`: Inspect the workflow subsystem
- `workflow_actions`: List workflow action definitions
- `workflow_run`: Execute a three-step local workflow and publish an artifact
- `workforce_health`: Inspect the workforce subsystem
- `workforce_agents`: List workforce agent descriptors

## FAQ

### Does it call real AI providers?

No. The server pins the gateway to a fake provider and does not authorise real calls.

### Which clients does it target?

Codex, Cursor and Cline.

### How is it run?

From a published OCI image over stdio, or from source with a local streamable HTTP option.

### Is it production ready?

The project calls itself a hardened public preview and says production still needs independent deployment evidence.

### Is Unified AI System MCP Gateway free?

Yes. Unified AI System MCP Gateway is open source, so you can use and self-host it for free.

### How do I add Unified AI System MCP Gateway to Claude Code?

Run this command in your terminal: claude mcp add unified-ai-system -- docker run -i --rm ghcr.io/happy520ai/unified-ai-system/mcp-server:0.8.0

### What are the best alternatives to Unified AI System MCP Gateway?

The closest alternatives to Unified AI System MCP Gateway 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/unified-ai-system-mcp-gateway
