Uber
Uber is a technology company headquartered in San Francisco, United States, primarily providing platform services for ride-hailing, food delivery, and freight. It connects passengers with drivers and consumers with merchants through a mobile application, operating across numerous countries and regions worldwide. The company originated with shared mobility as its core business and has since expanded into local lifestyle services, making it one of the largest mobility platforms globally.
Core Facts
Timeline (last 90 days)
According to reports by Forbes and The Information, companies such as Uber, Coinbase, Meta, and Microsoft quickly slowed token spending due to rising costs, and many organizations turned to internal or open-source models.
The initial fleet size, specific service area, and pricing of the London Robotaxi plan have not yet been announced.
Uber and Pony.ai plan to equip Robotaxi vehicles with safety drivers in the initial phase, and commit to extensive testing and establishing a safety compliance system.
Uber and Pony.ai are pushing forward with launching a Robotaxi service in London, and the two sides have now entered the final stage of regulatory approval and operational preparation.
Every MCP server and tool starts in a disabled state and must be explicitly reviewed and enabled by the responsible team.
The autocrawler tool reads interface definition files written in protobuf or thrift, has an LLM write friendly descriptions, and registers the tools, with each tool initially in a disabled state.
The gateway translates a single MCP call into one of three protocols — HTTP, gRPC, or t-channel — and then hands off execution to Uber's service mesh sidecar Motley.
In Uber's system, the registry is the control plane holding the tool catalog, and the gateway is the data plane handling actual invocations.
Uber internally has more than 5,000 MCP tools distributed across more than 800 servers.
Uber uses a large language model to automatically write agent-friendly descriptions for discovered tools.
40 more timeline events
All Facts (20)
Horovod is a data parallelism framework open-sourced by Uber.
65%VerifiedHorovod was originally open-sourced by Uber and adopts a Ring-AllReduce communication pattern.
65%UnverifiedKafka被Netflix、Uber、LinkedIn等公司大规模使用
90%UnverifiedGergely Orosz曾任职于Uber和微软等大型科技公司
90%UnverifiedUber拥有数千名工程师
85%UnverifiedShopify、Google（Submit Queue）和Uber（SubmitQueue）都有早期的合并队列实践
85%UnverifiedLinear由前Uber工程师Karri Saarinen于2019年创立
85%UnverifiedUber采用微服务架构，一个核心库的升级可能需要在数百个服务中同步修改
80%Partially VerifiedHorovod由Uber开源，基于Ring-AllReduce算法实现高效梯度同步
65%UnverifiedUber and Lyft faced similar growing pains during the same period as Airbnb's hyper-growth years
60%UnverifiedUber went public in 2019 with a market capitalization of approximately $82 billion
60%UnverifiedUber在2014-2016年高峰期每完成一笔订单平均补贴约40%的成本
60%UnverifiedAccording to reports by Forbes and The Information, companies such as Uber, Coinbase, Meta, and Microsoft quickly slowed token spending due to rising costs, and many organizations turned to internal or open-source models.
50%UnverifiedUber and Pony.ai plan to equip Robotaxi vehicles with safety drivers in the initial phase, and commit to extensive testing and establishing a safety compliance system.
50%UnverifiedThe initial fleet size, specific service area, and pricing of the London Robotaxi plan have not yet been announced.
50%UnverifiedEvery MCP server and tool starts in a disabled state and must be explicitly reviewed and enabled by the responsible team.
50%UnverifiedThe gateway enforces charter policies using Uber's internal access control system; policies are set per server and can differentiate by caller type (human, service, or agent), with authorization and PII redaction taking effect per tool.
50%UnverifiedThe gateway translates a single MCP call into one of three protocols — HTTP, gRPC, or t-channel — and then hands off execution to Uber's service mesh sidecar Motley.
50%UnverifiedUber internally has more than 5,000 MCP tools distributed across more than 800 servers.
50%UnverifiedUber uses a large language model to automatically write agent-friendly descriptions for discovered tools.
50%