As reported on November 18, 2026, Uber had built an MCP Gateway to give AI agents a consistent way to discover and call internal services. The platform supported more than 800 MCP servers and 5,000 tools, addressing fragmented integrations, duplicated infrastructure and inconsistent security practices.
Uber engineers described the gateway as a way to expose existing APIs as MCP tools without changing the underlying services. Its MCP Registry acts as a control plane, maintaining tool definitions, schemas, ownership and enablement settings. A separate Proxy Gateway handles execution, translating MCP requests into HTTP, gRPC or TChannel calls and converting the responses back into MCP-compatible results.
An automated component called AutoCrawler scans Uber’s internal interface-definition registries, extracts API definitions from Protocol Buffers and Thrift schemas, and generates MCP-compatible schemas. It uses a large language model to draft tool descriptions. Newly discovered tools remain disabled until their owning teams review and explicitly enable them. Configuration changes can reach the execution layer without a service restart or redeployment.
The gateway enforces access controls at both the server and tool levels through Uber’s internal system. It also supports redacting personally identifiable information and other sensitive data from tool responses. For integrations such as Jira and Google, a separate service exchanges internal user tokens for the appropriate third-party authentication tokens.
To avoid loading thousands of tool definitions into an agent’s context, Uber developed Omni MCP for incremental discovery: agents can search for servers, find relevant tools, retrieve individual schemas and call selected tools on demand. Response Projection lets agents request specific response fields rather than a full payload. Uber’s aifx command-line interface also lets coding agents discover and call tools, writing results to files for selective inspection.
Commenting on the architecture in LinkedIn posts, Microsoft CoreAI principal software engineer Ajay Sharma and Pipe17 president Kelly Goetsch compared it with established API gateway patterns. Tamir Dresher emphasized that security, discovery, observability and governance remain familiar distributed-systems challenges.
