MCP Switchboard
One governed endpoint that brings many MCP tool servers to Claude and ChatGPT, with an encrypted vault, per-tool policy, approvals and an audit log.
The problem
Connecting AI assistants to business tools means one configuration per tool per assistant, secrets scattered in config files, and no record of what the assistant did.
What we engineered
Put one switchboard in the middle: every tool behind one local endpoint, with its own permissions, and every call written down.
- A local aggregator that re-exposes many MCP servers through one endpoint.
- An AES-256-GCM vault, per-tool on or off and read, write or full policies, an approval gate and an audit log.
- Local OAuth for common providers and a generator that turns an OpenAPI spec into an MCP server.
How it connected
A dashboard, a command line and one-command install into several AI clients.
What technology was appropriate
- TypeScript on Node.
- Checked by a suite of verification scripts rather than screenshots.
Where it stands
Working alpha, released as v0.1.0 under Apache-2.0.
Current state, stated as it is today.
See it yourself
Automate operations
Workflow automation, AI agents and knowledge assistants that take repeated work off your team.
Security, evaluation and governance
Security reviews, evaluation, audit trails and human approval where an action matters.
Build software
Custom web apps, internal tools, SaaS products and the APIs that connect them.
Bring us the bottleneck.
If your problem looks like one of these, tell us. We will say whether a build is the right answer.
Bring us a bottleneck