# Zen MCP

> Multi-model orchestration — let Claude consult Gemini, GPT, and o-series as sub-agents.

[Canonical HTML page](https://top-mcps.com/mcp/zen-mcp) · [server.json](https://top-mcps.com/mcp/zen-mcp.json) · [methodology](https://top-mcps.com/about/methodology)

## Install

### Claude Desktop — `claude_desktop_config.json`

Paste under mcpServers. Fully quit and reopen Claude after editing.

```json
{
  "mcpServers": {
    "zen-mcp": {
      "command": "uvx",
      "args": [
        "zen-mcp-server"
      ],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### Claude Code — `CLI or .mcp.json`

Run from your repo. Commit .mcp.json to share with your team.

```shell
# export GEMINI_API_KEY=OPTIONAL
# export OPENAI_API_KEY=OPTIONAL
# export OPENROUTER_API_KEY=OPTIONAL
claude mcp add zen-mcp -- uvx zen-mcp-server
```

### Cursor — `.cursor/mcp.json`

Global path: ~/.cursor/mcp.json. Reload window after editing.

```json
{
  "mcpServers": {
    "zen-mcp": {
      "command": "uvx",
      "args": [
        "zen-mcp-server"
      ],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### VS Code — `.vscode/mcp.json`

VS Code uses the "servers" key (not "mcpServers").

```jsonc
{
  "servers": {
    "zen-mcp": {
      "command": "uvx",
      "args": [
        "zen-mcp-server"
      ],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### Windsurf — `~/.codeium/windsurf/mcp_config.json`

Open via Cascade → hammer icon → Configure.

```json
{
  "mcpServers": {
    "zen-mcp": {
      "command": "uvx",
      "args": [
        "zen-mcp-server"
      ],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### Cline — `cline_mcp_settings.json`

Open via the Cline sidebar → MCP Servers → Edit.

```json
{
  "mcpServers": {
    "zen-mcp": {
      "command": "uvx",
      "args": [
        "zen-mcp-server"
      ],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### Continue — `~/.continue/config.json`

Continue uses modelContextProtocolServers with a transport block.

```json
{
  "experimental": {
    "modelContextProtocolServers": [
      {
        "transport": {
          "type": "stdio",
          "command": "uvx",
          "args": [
            "zen-mcp-server"
          ],
          "env": {
            "GEMINI_API_KEY": "${GEMINI_API_KEY}",
            "OPENAI_API_KEY": "${OPENAI_API_KEY}",
            "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
          }
        }
      }
    ]
  }
}
```

### Codex CLI — `~/.codex/config.toml`

Codex uses TOML. Each server is a [mcp_servers.<name>] subtable.

```shell
# ~/.codex/config.toml
[mcp_servers.zen-mcp]
command = "uvx"
args = [
  "zen-mcp-server",
]
env = { GEMINI_API_KEY = "${GEMINI_API_KEY}", OPENAI_API_KEY = "${OPENAI_API_KEY}", OPENROUTER_API_KEY = "${OPENROUTER_API_KEY}" }
```

### Zed — `~/.config/zed/settings.json`

Zed calls them "context_servers". Settings live-reload on save.

```jsonc
{
  "context_servers": {
    "zen-mcp": {
      "command": {
        "path": "uvx",
        "args": [
          "zen-mcp-server"
        ]
      },
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
      }
    }
  }
}
```

### ChatGPT — `ChatGPT → Apps directory`

Zen MCP doesn't ship a hosted HTTPS endpoint today. ChatGPT supports remote MCP servers only — to use this server in ChatGPT you'll need to deploy it to a public HTTPS URL first (e.g. via Cloudflare Workers or Vercel) or wait for an official remote build.

```none

```

## At a glance

- **Maintainer:** Beehive Innovations (community)
- **Transport:** stdio
- **Auth model:** API key
- **Required secrets:** GEMINI_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY
- **Supported clients:** Claude, Claude Code, Cursor, Any MCP-compatible client
- **License:** Apache-2.0
- **Language:** Python
- **Latest version:** latest
- **Last verified:** 2026-05-26
- **GitHub stars:** 11,563 (fetched 2026-05-26T10:50:23.602Z)
- **Score:** 71/100 (rubric 2026-04 — see https://top-mcps.com/about/methodology)
- **Source:** https://github.com/BeehiveInnovations/zen-mcp-server

## Security & scope

- **Access scope:** network
- **Sandbox:** Calls outbound to whichever LLM providers you configure. Prompts and outputs flow through those providers under their privacy policies.
- **Gotchas:**
  - Multi-model calls amplify token spend — set per-provider budgets.
  - Provider-specific features (Anthropic prompt caching) do not pass through universally.

## Quick answer

**What it does.** Provides workflow tools that route to a chosen model, plus higher-level workflows (consensus across multiple models, deep thinking, codereview) that orchestrate multiple model calls inside one MCP tool.

**Best for:**
- Multi-model consensus
- Deep debugging
- Architecture decisions
- Cross-model code review
- Accessing specialized models inside Claude

**Not for:**
- Latency-sensitive work
- Routine single-step tasks

## Description

Zen MCP is an orchestration server that lets a primary model (typically Claude) delegate to other LLMs as sub-agents — Gemini, GPT, o-series, DeepSeek, local Ollama — through a unified `chat`, `thinkdeep`, `debug`, `analyze`, `codereview`, and `consensus` tool surface.

## Why it matters

Single-model thinking has known blind spots. Zen makes "ask three models, take the consensus" a single MCP call, which dramatically improves correctness on hard problems without leaving the primary chat.

## Key features

- Multi-model routing
- Consensus workflows
- Thinkdeep / debug / codereview / analyze tools
- Bring your own provider keys
- Local Ollama support

## FAQ

### Do I need keys for every provider?

No — set keys only for the models you plan to use. Zen can also route through OpenRouter to access many providers with a single key.

### How does consensus work in practice?

The `consensus` tool sends the same prompt to N configured models and asks them to converge on an answer, returning each model's response plus a synthesized one. Best on questions with a clear right answer.

### Does it work with local models?

Yes via Ollama and LM Studio. Useful for keeping sensitive prompts on-device while still doing multi-model work.

## Changelog

- **2026-05-26** — Refreshed install snippets and fact sheet; verified for 2026.
- **2025-02-20** — Initial directory listing.