Serve Robotics just made its sidewalk delivery robots something you can call from code. On November 11 the company launched the AV Delivery Developer Portal, a public API plus a sandbox that lets developers, businesses, hobbyists and even AI assistants request a robot pickup and dropoff without going through one of the big delivery apps.
For anyone building in robotics or agent software around the Peninsula, this is a useful signal. A robot fleet that used to sit behind third-party platforms is now exposed as an API with docs and a test environment, which looks a lot like the way cloud services and payments grew up. Below is what Serve announced, what the API does, where the limits are, and what founders and operators can take from it.
What Serve Robotics announced on November 11
According to The Robot Report, Serve released a public API that lets outside builders work with its autonomous sidewalk fleet, along with a sandbox where anyone can build and test applications before touching live robots. CEO and co-founder Ali Kashani framed it simply, saying people should not have to start a robotics company to build something useful with robots.
The same report describes Serve as a Redwood City company that spun out of Uber in 2021. It says Serve has deployed more than 2,000 robots across the U.S. and Australia, and that until now its robots were available only through third-party delivery partners. Earlier in 2026, Serve also acquired hospital service robot maker Diligent Robotics.
The developer launch follows a second move from the week before. On November 5, Serve introduced Physical AI Services, which opens the technology behind its robots (connectivity to off-board intelligence, training data curation and real-world foundation model data) to other companies building physical AI. Bedrock Robotics, which builds autonomous systems for heavy construction equipment, is listed among the launch customers. If you read our roundup of Bedrock, Serval and Hex funding, that name will ring a bell.
How the Serve Delivery API works for developers
The Serve Delivery API site describes the product as early access and says developers and AI agents can quote, dispatch and track sidewalk delivery robots with 3 REST calls. The flow starts with a quote request, then moves through dispatched, en route and delivered states, and the site lets you plan a sample trip by tapping 2 street corners on a map.
Serve handles the hard parts behind the call. The Robot Report says the company takes care of dispatch, routing, navigation and remote supervision, while the app or agent requests a pickup and dropoff, confirms details and tracks the robot until the delivery is complete. The sandbox lets builders simulate robots, run delivery workflows and rehearse edge cases first, and access to live robots is subject to Serve's approval.
The use cases Serve lists are deliberately small and local: a morning matcha, dry cleaning, a neighborhood package. For agent builders, this gives an AI assistant a real-world action it can take with a few lines of code.
The limits builders should plan around
Coverage is the first constraint to check. Per The Robot Report, deliveries are limited to Serve's active service areas and operating hours, and to items that fit inside the robot's cargo compartment. Anyone hoping to build a product on top of the fleet should check coverage in their target neighborhoods before writing a business plan around it.
The numbers also depend on which page you read. The Robot Report cites more than 2,000 robots deployed, while the developer site shows a counter of 1,505+ robots and 113+ neighborhoods. Treat those as company-reported figures that may measure different things, and confirm the live numbers in the docs when you scope a pilot.
What it means for Peninsula and South Bay robotics founders
The bigger story for local founders is the platform pattern. A robotics company with a working fleet is selling access to that fleet and to the stack behind it, which gives smaller teams a way to test ideas in the physical world without building hardware first. That pairs well with the trends in our breakdown of Q3 2026 robot orders, where demand for automation kept showing up across industries.
It also changes the build-versus-buy math for hardware teams. If your product needs last-mile movement but your edge is software, an API like this lets you prototype the experience while you decide whether custom hardware is worth it. Our guide to hardware prototyping in the South Bay covers the other side of that choice, and if you are building something worth recognizing, the TRR50 2027 nominations guide explains how to put your startup forward.
The Peninsula has a deep bench of people who have shipped robots, from Redwood City to Mountain View and Sunnyvale. That is a big reason Tribe organizes South Bay Renaissance, a series of in-person gatherings built to rebuild the South Bay and Peninsula startup and social scene one meetup at a time. Back on October 9, our Physical AI and Robotics Mixer at Circuit Launch in Mountain View brought builders in this space together in one room.
Questions to ask before you build on a robot fleet
If you are thinking about a product on top of Serve or a similar platform, a few practical questions come first. Ask whether the service area covers your users, what the approval process for live robots looks like, how pricing works once you leave the sandbox, and what happens to your customer experience when a robot is delayed or a delivery fails.
It is also worth thinking about dependency. Building on someone else's fleet is fast, but your product inherits their coverage, hours and policies. Teams that plan for a second provider, or keep the robot step behind a clean interface in their own code, will have an easier time adapting as more fleets open up.
Find your robotics people
Physical AI moves fastest when builders compare notes in person and online. Join the Futurists in Robotics Tribe to trade takes on fleets, APIs and what is shipping next, and follow the Tribe Events calendar on Luma for upcoming dinners, mixers and talks around the Peninsula and South Bay.
