home / mcp / praison ai mcp server

Praison AI MCP Server

Provides MCP access to PraisonAI tools via HTTP, STDIO, and WebSocket transports for seamless agent-to-agent communication.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "mervinpraison-praisonai": {
      "url": "https://api.example.com/mcp",
      "headers": {
        "GROQ_API_KEY": "your-key-here",
        "GOOGLE_API_KEY": "your-key-here",
        "OPENAI_API_KEY": "your-key-here",
        "OPENAI_BASE_URL": "http://localhost:11434/v1",
        "ANTHROPIC_API_KEY": "your-key-here"
      }
    }
  }
}

You can connect to and run an MCP server to enable inter-agent communication, tool hosting, and remote command execution for PraisonAI. This server supports multiple transports (HTTP, STDIO, and WebSocket) and allows you to expose tools or interact with remote MCP endpoints in a scalable, pluggable way. Use it to coordinate agents, share capabilities, and manage sessions across distributed clients.

How to use

Set up an MCP server to expose tools or consume MCP endpoints. You can run a local STDIO server that your PraisonAI agents connect to, or point your client at a remote HTTP MCP endpoint. Use the supported transports to match your deployment needs: local development (stdio), production hosting (http), or real-time bidirectional communication (WebSocket). Ensure you have the required API keys and environment configurations if you connect to external MCP servers.

How to install

Prerequisites: you need Python installed to run PraisonAI MCP components and, optionally, Node.js if you plan to use JavaScript tooling for MCP examples.

# Python prerequisites (recommended)
python3 --version
pip3 --version

# Install PraisonAI MCP components (example for CLI/SDK usage)
pip install praisonaiagents

MCP server configuration and connections

The MCP server can be connected via HTTP, STDIO, or WebSocket transports. Include environment variables shown below where needed. The following configurations illustrate typical usage from the source examples.

{
  "mcpServers": [
    {
      "type": "http",
      "name": "remote_mcp",
      "url": "https://api.example.com/mcp",
      "args": []
    },
    {
      "type": "stdio",
      "name": "memory_mcp",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"],
      "env": {"BRAVE_API_KEY": "YOUR_API_KEY"}
    },
    {
      "type": "stdio",
      "name": "brave_mcp",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {"BRAVE_API_KEY": "YOUR_KEY"}
    }
  ]
}

Security and safety considerations

When exposing tools or consuming MCP endpoints, apply origin validation, secure session handling, and, if available, token-based authentication for WebSocket and HTTP transports. Use human approvals for dangerous operations and enable guardrails where applicable.

Examples and tips

Expose a simple memory tool via STDIO for local testing, then switch to an HTTP MCP endpoint for production use. If you combine multiple MCP servers, ensure your client respects the first HTTP option as the default transport and fall back to STDIO or WebSocket as needed.

Notes

Keep MCP server configurations in version control, and document any environment-specific keys or endpoints separately. Regularly rotate API keys and review access controls to minimize exposure in shared environments.

Code snippets

For quick reference, you can reuse the following MCP server definitions as a starting point in your own configuration files.

Available tools

Tavily

Web search tool for fast information retrieval

DuckDuckGo

Web search tool with privacy-focused queries

Exa

Web search tool with advanced querying capabilities

Serper

Web search tool API integration

Jina

Web reader tool for extracting and reading content

Firecrawl

Web scraping tool for data extraction

Crawl4AI

Web scraping tool for structured data

Wikipedia

Knowledge source integration for Wikipedia content

ArXiv

Academic paper search and retrieval

HackerNews

News and discussion aggregator access tool

YouTube

Video content search and retrieval tool

Weather

Weather data retrieval tool

PostgreSQL

Database interaction tool for PostgreSQL queries

MySQL

Database interaction tool for MySQL queries

SQLite

Lightweight SQL database tool

MongoDB

NoSQL database interaction tool

Redis

Key-value store interaction tool

Qdrant

Vector store interaction tool for embeddings

GitHub

GitHub API integration for code and project data

Slack

Slack messaging tool integration

Discord

Discord messaging tool integration

Telegram bot tool integration

Email

Email send/receive helper tool

Notion

Notion workspace API integration

File

File system tool for reading/writing files

Shell

Shell command execution tool

Python

Python code execution tool for scripting

JSON

JSON data handling utility

CSV

CSV parsing and generation tool

Calculator

Mathematical calculator tool