Datadog, the cloud monitoring and security company, acquired Propolis, a New York startup that uses groups of AI browser agents to test web applications the way real users would. Datadog announced the deal in a post on its company blog on January 28, 2026.
Terms were not disclosed. Datadog did not say what it paid, how the deal was structured or when it closed.
Propolis was a two-person company at the time of the deal, according to its Y Combinator profile. Datadog said it was working to bring Propolis's features into its own platform and would share updates on that work later.
What Propolis built
Propolis went through Y Combinator's Spring 2025 batch and described its product as hands-off quality assurance (QA) run by browser agents. Customers gave it a web address and login details, and its agents explored the application to work out what real users try to do there, such as signing up, checking out or completing a workflow.
It then sent swarms of synthetic users through those journeys to check that they ended in the right result. Datadog said this approach suits applications whose behaviour is hard to script in advance, including complex, nondeterministic and agent-driven user flows.
Propolis's tests were also designed to repair themselves. When an application changed, later runs and continuous test generation updated the test suite so that coverage stayed current without engineers rewriting scripts by hand. Datadog said this can catch errors that traditional testing tools miss.
Why Datadog is buying
Datadog already collects large amounts of production data from its customers' software, including distributed traces, logs and Real User Monitoring (RUM) sessions. Its plan is to feed that data to Propolis's agents, so they can infer what users actually do from live usage rather than relying only on how designers expected the product to be used.
The company said the result would automate testing from end to end and turn quality checks from a slow gate before each release into a continuous signal inside CI/CD pipelines and in production.
In principle, that could let the agents prioritise the journeys that customers use most, and test new paths as soon as real users start taking them, instead of waiting for a QA engineer to notice and write a test. Datadog positioned the combination as a way to catch breakages in flows that are hard to predict, such as those involving generative AI features.
The blog post announcing the deal was written by Hugo Kaczmarek and Ruxanda Lueck of Datadog.
The founders
Propolis was founded in 2025 by Marc Papazian, its chief executive, and Matt White, its chief technology officer. Papazian previously worked as a deployment lead at Palantir and later as director of product at an AI startup. White was a senior engineer at Airtable, where he worked on infrastructure meant to reduce manual QA work.
The company's funding was not disclosed beyond its participation in Y Combinator, which invests in the startups in each batch. Datadog did not say what roles the founders will take.
AI agents move into software testing
Software testing has traditionally depended on engineers writing and maintaining scripts that break whenever an interface changes. Propolis's pitch was that agents could take over both jobs: discovering what to test and keeping the tests working as the product evolves.
That problem grows as more applications include AI features whose output varies from run to run, making it harder to check results against fixed expected values. Database Trends and Applications, reporting on the deal, highlighted Propolis's focus on goal- and output-oriented testing for these kinds of journeys.
For Datadog, pairing an agent-based tester with production monitoring data is an attempt to close the loop between what breaks in production and what gets tested before release. The company has not said when Propolis's technology will be available to its customers or how it will be priced.
