The enterprise agent surface just grew to 5,000 tools at one company. uber's mcp gateway runs 800+…
the enterprise agent surface just grew to 5,000 tools at one company. uber's mcp gateway runs 800+ mcp servers and exposes 5,000 tools with every one starting off, routed through a registry that translates mcp into http, grpc and tchannel.
the bottleneck moved from building tools to deciding which ones turn on.
Context
A summary of Uber Engineering's October 1, 2026 post says Uber's MCP Gateway hosts over 800 MCP servers and over 5,000 tools, and sits between AI agents and both existing back-end services and native MCP servers. It is split into a registry, which is the catalog and the source of truth for discovery, ownership and enablement, and a proxy that translates MCP calls into HTTP, gRPC or TChannel and converts the responses back.
The summary says an automated crawler registers tools for new services and APIs disabled by default, and that every server and tool starts disabled. The owning team has to review and enable them before they can be called, and any change to a tool description needs the owner's approval. Authorization is applied at tool level using Uber's access control system, and responses are redacted for sensitive data.
To keep agents from loading hundreds of schemas, the gateway offers one proxy with four tools (discover a server, discover its tools, get one tool's schema, invoke a tool), and coding agents use a command line tool with list, search and call commands. A second write-up notes the scale figures are Uber-reported and that this is an internal architecture, not a gateway release.
The note's figures (800+ servers, 5,000 tools, every one starting off, a registry translating MCP into HTTP, gRPC and TChannel) all match the summaries read. Both summaries rest on Uber's own report, and the original Uber page was not opened here, so the numbers are Uber's own and are not independently checked.
The note says the surface 'just grew' to 5,000 tools. The sources give a count at the time of the October 1 post and no earlier number, so how fast it grew is not stated.
That the bottleneck moved from building tools to deciding which ones turn on is the author's opinion. The sources do describe owner approval as the gate, which is consistent with it, though they also describe an automated crawler that produces the tool definitions.
Related work
- Uber's MCP Gateway hosts over 800 servers and over 5,000 tools, off until owners enable them (subagentic.ai, summarizing Uber Engineering's October 1, 2026 post) ↗Source for the registry and proxy design, default-off enablement and the discovery tools.
- Uber details a governed MCP gateway for more than 800 servers and 5,000 tools (Rise Productive) ↗Second summary that states the figures are Uber-reported and the design is internal.
Watch next
- Open Uber Engineering's own post to confirm the figures and wording. Look for how many of the 5,000 tools are currently enabled.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 8 October 2026 at 22:48 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →