Toolport lets you run unlimited MCP servers without token bloat
Keep 15+ MCP servers active with zero context drain and built-in security.
Toolport is a new open-source tool (MIT license) from developer tsouth89 that solves two major pain points for AI agent users: context window bloat and security risks. When you attach multiple MCP (Model Context Protocol) servers to an agent, each server typically sends its tool descriptions every turn, consuming precious context tokens. Toolport eliminates that by managing tool availability efficiently — context usage stays flat whether you have 1 or 15 servers on call. This means you can keep a full arsenal of tools (web search, code execution, file access, etc.) ready without having to manually toggle servers on and off to save tokens.
Beyond context savings, Toolport adds security layers that are rare in this space. It actively monitors for rug-pull attacks (malicious tool replacements), tool poisoning (injected harmful instructions), and agentjacking (agents being hijacked to unauthorized endpoints). All API keys and secrets are stored in the OS keychain, not in plaintext config files. The tool also lets you import your existing Claude Desktop MCP configuration and re-authenticate once — those servers then become available in Claude, Cursor, and any of the 20+ supported AI agents. For developers who constantly switch between clients, this eliminates repetitive setup work.
- Context usage stays constant regardless of number of active MCP servers (tested with 15+)
- Built-in security detection for rug-pull, tool poisoning, and agentjacking attempts
- Import existing Claude MCP configs; automatically connects to 20+ AI agents including Cursor
Why It Matters
For developers juggling multiple AI agents, Toolport eliminates context waste and security risks in one lightweight open-source tool.