Labelbox

Labelbox is a platform that builds training environments and expert data for frontier AI labs and, through a product called Recursion, provides managed AI agents for engineering, research and business teams. The company is known for data labeling, and its current site is centered on these two offerings. Recursion agents start on their own from triggers such as a schedule, an alert, a failed build, a new checkpoint, a message or an API call, then work in the background until the job is done. Example agents include a Feature Builder that turns ready tickets into tested pull requests, a Bug Fixer that reproduces new issues and writes a fix with a regression test, and a Dependency Upgrader that opens one mergeable pull request per package. Agents connect to more than 100 apps, including GitHub, Jira, Slack, Datadog, PagerDuty, Salesforce, Zendesk and NetSuite. Work is organized as coordinated sessions in which specialist agents run in parallel, each run is graded by a grader agent, and graded sessions leave notes and skills that later sessions read. Use cases extend beyond engineering to ML research, evals, product, finance, sales and procurement. Labelbox serves enterprises that want to hand off recurring work to agents, and frontier AI labs that need environments and expert data to train models. Pricing is not published on the site. Access is through a Try Recursion button or by contacting the company.