Conduit launches local MCP gateway

Conduit addresses the massive "context tax" of MCP by lazy-loading tool definitions, turning a major bottleneck for AI agents into a non-issue.

• –Exposes three meta-tools instead of hundreds, reducing tool overhead by 97% per request and total token usage by ~90%.
• –Improves security by storing API credentials directly in the OS keychain and injecting them at runtime rather than in plaintext config files.
• –Offers per-tool governance with live toggles to disable destructive commands across all connected AI clients instantly.
• –Runs entirely locally as a native desktop application with live observability features like latency metrics and call audit trails.
• –Simplifies multi-client setups, allowing developers to configure servers once and route them to Claude, Cursor, VS Code, and others.