How we build a governed, multi-tenant MCP hosting platform — plus what shipped recently. For a plain changelog of releases, see the changelog.
A plain explanation of the protocol, the problem it replaces, and what a governed MCP server adds that the spec itself doesn't.
What actually happens when a spec becomes a tool: the mapping rules, the risk classification, and how to run it in minutes.
Your Swagger 2.0 doc doesn't need to become OpenAPI 3.x first — paste it in as-is and get a live MCP endpoint.
A hands-on walkthrough using a public demo store API: paste an OpenAPI spec, generate risk-classified tools, publish, and call it live from Claude.
Publish once, connect from claude.ai, Claude Desktop, or Claude Code — the same install URL works everywhere Claude runs.
The same published MCP server you'd hand to Claude works as a ChatGPT connector too — one build, no per-client integration.
Click through the console, or just ask an AI client to do it for you — both run the same pipeline, zero code either way.
No code change to your API. Start with one low-risk endpoint in front of it and expand the governed surface from there.
Why database-enforced Row-Level Security — not app-level WHERE filtering — is the right primitive for multi-tenant isolation, and how it's wired into every tenant-owned table on this platform.
Hands-on, end-to-end walkthroughs if you'd rather build something than read about the internals.
From a public OpenAPI spec to a governed, published MCP server that Claude can call — sign-up to first tool call.
How the platform's own Postgres database enforces per-tenant isolation at the row level — the tutorial version of the deep-dive above.
Trust boundaries (internal/partner/public) and per-tool risk levels — how to expose an internal API to an LLM safely.