E2e Raises $8M Seed to Verify Code Written by AI Agents

E2e has raised $8 million in seed funding to build a verification layer for software written by coding agents. The company, which changed its name from TesterArmy, announced the round on November 3, 2026. Backers include Credo Ventures, KAYA, Point Nine, Baobab Ventures and Script Capital, with individual investments from Snowflake co-founder Marcin Żukowski and Uber co-founder Oscar Salazar.

The rebrand signals a wider ambition. Testing is the first piece of a broader verification strategy meant to support agents throughout their work.

Why did TesterArmy become E2e?

The team started TesterArmy earlier in 2026 and went through Y Combinator. The closed platform blocked customers in three ways. Users could not store tests alongside their code, could not choose any model, and could not run tests locally next to a coding agent.

Those limits shaped the new product. Tests remain code owned by the user. Model selection stays with the user. Tests can run locally or in continuous integration.

The company launched its open-source framework on October 1, 2026. The rebrand extends the framework's name to the whole company after the release became the product developers associated with the team.

How does the framework cut model costs?

The TypeScript framework supports web, iOS and Android through one API. Developers use role-based locators and auto-retrying assertions where precision matters. For looser goals, they add agent.act(), agent.assert() and agent.extract() steps, where describing an outcome beats scripting every action.

The cost mechanism sits in caching. When an agent step passes with a recorded check, e2e caches its actions and replays them on later runs without another model call. If an interface changes enough to break the replay, the live agent takes over and finds a new path. Model usage therefore follows changes to the interface, not every repetition of an unchanged test.

Agent steps can run on AI SDK models, including local models, or through existing ChatGPT, GitHub Copilot and SuperGrok subscriptions. An agent skill and an MCP server connect the framework to Claude Code, Cursor and Codex, so those coding agents can explore an app, write tests using interface locators and read failure reports.

What does adoption look like?

E2e reports 1.5 million npm downloads and 12,000 GitHub stars in the framework's first month. Sixty external contributors have joined the project. Daily downloads run at roughly 100,000.

The framework is available on npm under an Apache-2.0 license, but it remains pre-1.0. The project repository warns that APIs and configuration can change between minor releases. Desktop applications are not yet covered.

What comes next?

The company plans to launch e2e cloud in 2027. Teams will run the framework on managed infrastructure with a range of models and devices. Additional verification tools are planned afterward. The open-source framework stays Apache-2.0.

The enterprise offering keeps the framework free for enterprise use and charges for engineers who help with rollout or custom work. Those engagements cover test migration, continuous-integration setup, stack-specific development and handover. Tests and applications run on customer-controlled infrastructure.

Existing TesterArmy customers can use the legacy platform until January 15, 2027. The company will help move their tests to e2e before then. It is also hiring in Warsaw and San Francisco as it expands the team.

The seed round funds a shift from a closed testing service into an open verification business with reported early traction. The execution challenge: build enterprise and cloud operations around a framework that is still pre-1.0, while preserving the local control, model choice and open licensing that distinguish it.