Convex MCP Server

convexbackenddatabaseserverlessdeveloper-tools

Summary

Ships inside the Convex CLI — lets an agent inspect deployments, read table schemas, run read-only queries, call functions and check health insights.

Features

Lists table schemas and paginates real documents so agents stop guessing at data shapes
Runs read-only queries against a live deployment
Enumerates and executes deployed Convex functions with arguments
Surfaces 72 hours of health metrics and performance issues
Reads and manages deployment environment variables
Production deployments blocked by default behind an explicit opt-in flag

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

Open your agent's MCP configuration file.
Merge the snippet below into it.
Restart the agent — the "Convex MCP Server" tools will be available.

{
  "mcpServers": {
    "convex": {
      "command": "npx",
      "args": [
        "-y",
        "convex@latest",
        "mcp",
        "start"
      ]
    }
  }
}

Claude Code

Add the server with a single CLI command.

claude mcp add convex -- npx -y convex@latest mcp start

Claude Desktop

Add the server to claude_desktop_config.json (Settings → Developer → Edit Config).

{
  "mcpServers": {
    "convex": {
      "command": "npx",
      "args": [
        "-y",
        "convex@latest",
        "mcp",
        "start"
      ]
    }
  }
}

Cursor

Installing the Convex plugin for Cursor starts the server automatically.

{
  "command": "npx",
  "args": [
    "-y",
    "convex@latest",
    "mcp",
    "start"
  ]
}

Pin the server to a specific project directory. Add --preview-name to target a preview deployment.

{
  "command": "npx",
  "args": [
    "-y",
    "convex@latest",
    "mcp",
    "start",
    "--project-dir",
    "/path/to/project"
  ]
}

Description

The Convex MCP server is part of the Convex CLI rather than a separate package, so any project with Convex installed already has it. It gives an AI agent a live view of the backend it is writing code against, which closes the usual gap where the agent can read your source but has no idea what the deployed data actually looks like.

Tools, by area. Deployment — a status tool to find and select which deployment to work against. Tables — list schemas, paginate through documents, and run read-only queries so the agent can check its assumptions against real rows instead of guessing at shapes. Functions — enumerate deployed functions and execute them with arguments. Insights — surface health metrics and performance problems observed over the past 72 hours. Environment variables — read and manage deployment configuration.

The insights surface is the one that changes how an agent behaves: it can be pointed at a slow or failing deployment and diagnose from the platform's own telemetry rather than from a stack trace someone pasted in.

Safety default worth knowing. Production deployments are blocked out of the box. The --dangerously-enable-production-deployments flag exists but grants full read/write access to production data, and the name is a fair description of the risk. Development and preview deployments are targeted with --project-dir, --prod and --preview-name flags.

Setup. Add npx -y convex@latest mcp start to your MCP client config. If you use Claude Code or Cursor, installing the Convex plugin starts the server for you, so there is nothing to configure by hand.

Related MCP Servers

MCP: GitLab MCP Server

by GitLab

New

GitLab's own MCP server — issues, merge requests, CI/CD pipelines, job logs, wikis and semantic code search over an OAuth-authorised HTTP endpoint.

Developer Tools

MCP: Mobile MCP

by Mobile Next

New

Platform-agnostic MCP server that drives iOS and Android apps on simulators, emulators and real devices through the accessibility tree rather than screenshots.

Developer Tools