Skip to main content

One Model Context Protocol (MCP) server for AI agents to handle thousands of tools

Project description

image

Overview

Strata is a unified MCP server that intelligently manages thousands of tools across multiple applications, presenting them to AI agents progressively without overwhelming the model context.

[!TIP] 🚀 Looking for production-ready solution? We offer a fully managed Strata service with enterprise-grade performance, access to more features and advanced AI model optimizations. Get started instantly →

This repository contains our open-source implementation - a streamlined version you can easily deploy as your MCP aggregation manager in your local infrastructure.

🚀 The Problem We Solve

Imagine you have:

  • 📦 Dozens of MCP servers, each with dozens of tools
  • 🤖 AI models seeing 100s-1000s of tools at once
  • 📉 Degraded performance due to massive context

✨ The Strata Solution

Instead of flooding your AI model with hundreds of tools, Strata acts as an intelligent router:

🔍 Smart Discovery → Only exposes a few essential tools to the model
🎯 Progressive Access → Finds and surfaces the right tool when needed
Optimized Context → Maintains peak model performance
🔗 Seamless Integration → Works with your existing MCP servers, and easy configuration

Your AI agent gets access to thousands of tools while maintaining high performance! ⚡

Quick Start

Installation

pipx install strata-mcp

Or with pip:

pip install strata-mcp

For development:

pip install -e .

Configure MCP Servers

You can configure your MCP servers by strata CLI tool. Or manually configure it in a JSON file, just like your other MCP JSON config files.

Add MCP servers

  • Stdio Server
strata add --type stdio <server_name> npx @playwright/mcp@latest
  • SSE Server
strata add --type sse <server_name> http://localhost:8080/mcp/ --env API_KEY=your_key
  • HTTP Server
strata add --type http <server_name> https://api.githubcopilot.com/mcp/ --header "Authorization=Bearer token"
  • HTTP Server with OAuth
strata add --type http <server_name> https://mcp.notion.com/mcp --auth_type oauth

List Servers

strata list

Enable/Disable Servers

strata enable <server_name>
strata disable <server_name>

Remove Servers

strata remove server-name

Manual Configuration

Configuration is stored in ~/.config/strata/servers.json by default. You can specify a custom config path:

strata --config-path /path/to/config.json add --type stdio ...

Config Format

{
  "mcp": {
    "servers": {
      "github": {
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-github"],
        "env": {
          "GITHUB_TOKEN": "your_token"
        },
        "enabled": true
      },
      "api-server": {
        "type": "http",
        "url": "https://api.example.com/mcp",
        "headers": {
          "Authorization": "Bearer token"
        },
        "enabled": true
      }
    }
  }
}

Environment Variables

  • MCP_CONFIG_PATH - Custom config file path
  • MCP_ROUTER_PORT - Default port for HTTP/SSE server (default: 8080)

Running Strata MCP servers

Strata itself is a MCP server. You can run and use it like a normal MCP, in stdio mode or http/sse mode.

Stdio Mode (Default)

Run without arguments to start in stdio mode for direct MCP communication:

python -m strata
# or
strata

HTTP/SSE Server Mode

Run with port to start as HTTP/SSE server:

strata run --port 8080

Tool Integration

Strata can automatically configure itself in various AI assistants and IDEs that support MCP.

Add Strata to Claude Code

# Add to user configuration (default)
strata tool add claude

# Add to project-specific configuration
strata tool add claude --scope project

Add Strata to Gemini

strata tool add gemini

Add Strata to VSCode

strata tool add vscode

Add Strata to Cursor

# Add to user configuration (~/.cursor/mcp.json)
strata tool add cursor --scope user

# Add to project configuration (.cursor/mcp.json)
strata tool add cursor --scope project

Supported scopes:

  • user: Global configuration (default)
  • project: Project-specific configuration
  • local: Same as project (for Cursor)

Note: VSCode doesn't support scope parameter and will use its default behavior.

Available Tools

When running as a router, the following tools are exposed:

  • discover_server_actions - Discover available actions from configured servers
  • get_action_details - Get detailed information about a specific action
  • execute_action - Execute an action on a target server
  • search_documentation - Search server documentation
  • handle_auth_failure - Handle authentication issues

Development

Running Tests

pytest

Project Structure

  • src/strata/ - Main source code
    • cli.py - Command-line interface
    • server.py - Server implementation (stdio/HTTP/SSE)
    • tools.py - Tool implementations
    • mcp_client_manager.py - MCP client management
    • config.py - Configuration management

Examples

Running GitHub MCP Server through Router

# Add GitHub server (official HTTP server)
strata add --type http github https://api.githubcopilot.com/mcp/

# Run router in stdio mode
strata

# Or run as HTTP server
strata run --port 8080

Running Multiple Servers

# Add multiple servers
strata add --type stdio playwright npx @playwright/mcp@latest
strata add --type http github https://api.githubcopilot.com/mcp/

# List all servers
strata list

# Run router with all enabled servers
strata run --port 8080

Ready to experience the power of thousands of AI tools?

👉 Get Instant Access at Klavis AI (YC X25) 👈

From zero to thousands of tools in under a minute! ⏱️

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

strata_mcp-1.0.1.tar.gz (114.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

strata_mcp-1.0.1-py3-none-any.whl (43.3 kB view details)

Uploaded Python 3

File details

Details for the file strata_mcp-1.0.1.tar.gz.

File metadata

  • Download URL: strata_mcp-1.0.1.tar.gz
  • Upload date:
  • Size: 114.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.19

File hashes

Hashes for strata_mcp-1.0.1.tar.gz
Algorithm Hash digest
SHA256 829507c47a2b790971c2b7928d82cd09ef4654633a03b9c8bdf51160de30487c
MD5 c9642e2dcddd964e831e70a59a0adf1f
BLAKE2b-256 86cb5b37db5e5019765e5eed39fccd53a67acf40de27ef706d0e55b4b795cfcb

See more details on using hashes here.

File details

Details for the file strata_mcp-1.0.1-py3-none-any.whl.

File metadata

  • Download URL: strata_mcp-1.0.1-py3-none-any.whl
  • Upload date:
  • Size: 43.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.19

File hashes

Hashes for strata_mcp-1.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 da69b1989e718067c1e3fa84756dc6bb3cbe306681cdea803b1b3cd5063f641d
MD5 b2204c641ed2e0c9b2c44a386fe8e5b6
BLAKE2b-256 71d8224db6396db5acaa59810381994eb18f6ed73d9a3ce7828b7c74ecf83330

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page