Every MCP server dumps its whole tool list into your agent's context on every request. 3 servers cost ~24k tokens before you even say hi. Conduit routes them through one local gateway that exposes 3 meta-tools the agent searches on demand. Measured: 97% less tool overhead per request, ~90% fewer tokens, same task success. Works on one AI tool or five, cloud or local. Keys in your OS keychain, live toggles, no cloud, no account. Free and open source.
Conduit
