Toolport
Toolport is a free, open-source local MCP (Model Context Protocol) gateway that consolidates all your MCP server configurations into a single, secure application. Instead of pasting the same server settings and API keys into every AI client, you set up servers once in Toolport, and they become available to all supported clients like Claude, Cursor, Codex, and more, with hot toggles and no restarts. The tool addresses the token overhead problem inherent in MCP, where each server dumps its full list of tool definitions into the AI's context window, consuming tokens even before any user input. Toolport collapses hundreds of tools into a handful of meta-tools that the agent can search on demand, reducing token usage by up to 91% while maintaining task success rates, as shown in benchmarks. Security is a core feature: Toolport runs entirely on your machine, stores API keys in the OS keychain (never in client configs or the cloud), and fingerprints every tool definition to detect rug pulls (definition changes) and tool poisoning (hidden instructions). It provides per-tool governance, letting you toggle off destructive or sensitive tools fleet-wide, along with live observability metrics like latency, error rates, and a full audit trail. For teams, Toolport offers a centralized version with shared server configurations, central policy management, exportable audit trails, and the same token savings at scale. It can be self-hosted for compliance and data sovereignty, with a free tier for up to 5 people.
- Category
- AI Agents
- Pricing
- Free
Problem solved
Managing MCP server configurations and API keys across multiple AI coding assistants leads to repetitive setup, security risks from exposing secrets in plaintext configs, and excessive token consumption from loading full tool definitions in every context, degrading agent performance and increasing costs.
Use cases
- Unify MCP server management across Claude, Cursor, Codex, and other AI clients
- Reduce token usage and context pollution by using lazy tool discovery
- Enforce security policies and prevent tool tampering via fingerprinting and integrity checks
- Centralize API key storage in OS keychain instead of per-client config files
- Monitor and audit all tool calls across your AI agents
- Roll out governed MCP server access for development teams
Audience
- Developers
- ML Engineers
- Data Scientists
- Enterprise
- Startups
Pros
- Reduces token consumption by up to 91% without sacrificing task success
- Single configuration point for sharing MCP servers across all AI clients
- Local-first security with tool integrity checks and OS keychain integration
- Seamless lazy discovery of tools, keeping context flat even with many servers
- Built-in observability and per-tool governance for teams
Cons
- Requires manual installation and local resource management
- Only works with MCP-compatible clients and servers
- Team features self-hosting may add complexity for non-technical users
- Limited community and documentation as a relatively new tool
- Desktop-only, no web-based management interface
Official website: https://toolport.app/
Canonical LambdaBase page: https://www.lambdabase.com/tools/toolport