Building what software development is becoming
At Uber, DragonCrawl showed us that AI could operate real software and improve the next run. Revyl is building that loop for every team.
Why Revyl
Agents write code they never see run
A coding agent can open a pull request in minutes, then waits on a person to build it, run it and check it. Revyl gives the agent a runtime to do that itself.
- Build the change
- On RevylBuilds the branch on Revyl’s cloud runners
- On their ownHands the branch to a person to build
- See it running
- On RevylOpens it on a live device, browser or Mac
- On their ownNever sees the app it changed
- Prove it works
- On RevylPosts proof with screenshots on the pull request
- On their ownAsks a reviewer to check by hand
- Remember what it saw
- On RevylAdds each screen to the record in Atlas
- On their ownStarts from scratch on the next task
RevylAgents build, run and verify every change on the runtime it ships to, and what they see becomes your record.
Our story
How a research project became the runtime for cloud agents
origin.trace
- ubc.vancouvertwo friends at UBC
- uber.dragoncrawlAI tests Uber’s app
- research.publishedthe industry asks for it
- revyl.foundedRevyl founded
- yc.f24F24
- platform.compoundsbuild · develop · verify · atlasbuild · develop · verify · atlas
- 01dragoncrawl()Uber · 2023AI learns to test a real appAt Uber, Anam helps build DragonCrawl, an LLM-driven framework that tests the mobile app the way a person would. It catches eleven critical incidents and saves roughly $25 million in its first four months.
- 02found()2024The research escapes the labUber publishes the work, and teams everywhere want the same capability. Landseer and Anam, friends from UBC and both Canadian, start Revyl in early 2024 to build it for everyone.
- 03accelerate()YC F24Y Combinator, Fall 2024Revyl moves to San Francisco for YC’s Fall 2024 batch, with a pre-seed led by Canada’s Panache Ventures. Canadian original, Valley velocity.
- 04compound()TodayTesting became the runtimeWhat started as AI testing grew into the runtime for cloud agents. Agents build, run and verify every change on Revyl’s devices and Macs, and what they see becomes the record of how the product behaves.
What we believe
Everything we build turns intent into verified software
Runtime over assumption
Ground every decision in the behavior of the real application.
Evidence over ceremony
Make it easy to see why a change works and where it failed.
Compounding context
Ensure every run makes the next person or agent more capable.
Backed by
We started in Canada and build in San Francisco
Y CombinatorF24
Investor
Investor
Led pre-seed · Canada
Feld VenturesInvestor- AngelsOperators from Uber, Meta, and Nvidia, including a co-creator of OpenTelemetry.
Careers
Open roles on the founding team
- Staff Infrastructure Engineer — Device CloudOwn the technical strategy and systems that scale the device cloud and distributed compute platform behind Revyl.
- Team
- Engineering, Infrastructure
- Location
- San Francisco, CA, US · Remote (US)
- Type
- Full-time
- Software EngineerBuild the AI agent loop, device infrastructure, workflow orchestration, and product surfaces of the platform.
- Team
- Engineering, Full stack
- Location
- San Francisco, CA, US
- Type
- Full-time
- Engineering InternShip production work across agent systems, device and Mac infrastructure, and product with high ownership alongside the founders.
- Team
- Engineering, Full stack
- Location
- San Francisco, CA, US
- Type
- Internship
- GTM InternCreate content, talk to customers, and ship projects that grow the business alongside the founders.
- Team
- Growth and marketing
- Location
- San Francisco, CA, US
- Type
- Internship
Come build it with us
We are hiring engineers who want to own the loop from code to running, verified software.