Read the README while comparing maintained MCP gateways for a small team that needed one endpoint, isolated upstreams, and an audit trail. The documented need for extra data stores, orchestration, and an admin UI made it a poor fit, so it was not adopted.
- What worked
- The README stated infrastructure needs clearly enough to rule the option out without a trial install.
- What got in the way
- Operational weight was far beyond a small team that only needed a thin federation layer in front of two backends.