Conduit cuts agent tool tokens by hiding long MCP tool lists
Conduit is a that lets many AI apps share the same MCP tool servers. Claude, Cursor, Codex, and similar clients can point to Conduit instead of each app being configured separately. Its main cost-saving feature is .
Instead of sending every tool description to the AI on every request, Conduit first exposes only three meta-tools for status, search, and tool calls. In the example given, 3 with 62 tools would add about 24,000 tokens before any real question is asked, while Conduit lowers that to about 660 tokens. Its benchmark claims up to 91% fewer total tokens at the same task success rate, 97% less tool-description overhead per request, and 99.6% less overhead on a real 415-.
Secrets are stored in the OS keychain rather than client , and Conduit also adds tool-change detection, quarantine warnings, detection, tool toggles, and call logs.
Key points
- Conduit lets multiple AI clients share one instead of separate tool setups.
- shows three meta-tools first instead of every tool description.
- The example reduces tool-definition cost from about 24,000 tokens to about 660 tokens.
- The project claims up to 91% fewer total tokens with the same task success rate.
- Secrets stay in the OS keychain, and the gateway watches for risky tool changes and .