MCP Servers

Connect external tools to your AI agents. Filesystem. Search. Databases. GitHub. Slack.

No account neededNo API key for first MCPRuns locally

Available MCP Servers

No packages found

Try a different search term or adjust your filters.

What is an MCP Server?

An MCP server is a tool connector that lets AI agents interact with files, databases, APIs, developer tools, and other systems. AgentNode helps you discover, install, and run them.

Your First MCP in 3 Steps

Install AgentNode

bash

pip install agentnode-sdk

Python 3.10+ required. No account needed.

Install an MCP Server

bash

agentnode install mcp-filesystem

Connects your agent to the local filesystem. No API key needed.

Check & Run

bash

agentnode mcp doctor mcp-filesystem
agentnode run mcp-filesystem --input '{"path": "."}'

Doctor confirms your setup works. Run executes the MCP server and returns results.

Expected output

{"content": [{"text": "README.md\nsrc/\ntests/"}]}

That's it. Your first MCP server is running.

Next: Add an API-powered MCP

Now try an MCP that connects to an external service. This one searches the web via Brave Search.

bash

agentnode install mcp-brave-search

Set your API key

export BRAVE_API_KEY="your-key-here"

Get a free Brave API key

bash

agentnode mcp doctor mcp-brave-search
agentnode run mcp-brave-search --input '{"query": "latest AI news"}'

Publish your own MCP server

Maintain an MCP server? List it in the AgentNode catalog. Verify the manifest locally, then submit it for review — and track the submission status from the CLI.

bash

agentnode mcp verify .       # check your agentnode.yaml
agentnode mcp submit .       # submit for catalog review
agentnode mcp status <id>    # track review status

Submissions are reviewed before they go live. Prefer the browser? Submit via the web form — or see the CLI reference for the full publishing flow.