Tusk
Tusk AI, Inc. · Coding · Productivity
Tusk is an AI testing agent that writes and runs unit and API tests for pull requests, then flags the bugs it finds. Tusk AI, Inc. shut the product down on June 10, 2026.
About Tusk
What Is Tusk
Tusk began in 2024 as a Y Combinator W24 company with a narrow promise. Send the bug reports your customers file to an AI agent, get a pull request back. Small engineering teams lose whole afternoons to fixes that take ten lines of code, and Tusk existed to take those tickets off their hands.
By 2026 the product had grown into an AI testing agent. It generated unit tests and API tests for every pull request, ran those tests on its own, and iterated until they passed. Testing is the part of a release that nobody has time for, and Tusk sold itself as the teammate that does it without being asked.
The limits matter if you're evaluating anything it left behind. Tusk needed a GitHub repository, an existing test setup, and enough compute tolerance to let the agent retry a failing test several times. Its vendor claims about catching edge cases came from Tusk's own benchmarks, not from independent review.
The bigger limit is the calendar. Tusk AI, Inc. announced on June 11, 2026 that the product had shut down the day before. The GitHub App was removed from customer organizations, Tusk Drift users were told to uninstall the SDK and delete the CI workflow, and support ran for one month after. The company pointed users toward Cognition's Devin for test automation and code review.
Getting Started
- Create a Tusk account and install the GitHub App on the repositories you want it to read and write.
- Install the CLI in each service root, which took one command, then run the setup step so Tusk could index your test patterns.
- Open a pull request as usual. Tusk posted generated tests as a comment, ran them in your CI, and kept revising the ones that failed.
- Review the tests like any other code. Tusk reported that about 69% of the tests it generated ended up merged into pull requests.
Product Information
A quick look at Tusk's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Small engineering teams without a dedicated QA hire
- Startup founders who still review code
- Platform and backend engineers
Tasks
- Raising code coverage before a launch
- Catching regressions in pull requests
- Watching third party API contracts
Scenarios
- A legacy codebase with low coverage and no budget for a rewrite
- A team shipping several times a day across three or four repositories
- Due diligence on tools that no longer exist
Key features
Unit Tests Written for Every Pull Request
Tusk read the diff, pulled in context from your existing tests, and proposed unit tests as a comment on the pull request. The agent ran what it wrote and retried the ones that failed, so the output was executable rather than a suggestion. For teams that treated testing as a chore, getting working tests without opening an editor was the whole value.
API Tests Built From Real Traffic
Tusk Drift recorded live traffic against your endpoints and turned it into API tests. Instead of writing mocks, you got tests derived from what your service actually received. That mattered for API contract monitoring, where a silently changed response breaks callers you never hear from until a customer complains.
CoverBot Backfills Coverage on a Schedule
CoverBot worked like Dependabot, except the pull requests it opened were full of unit tests instead of dependency bumps. Point it at a repository, pick daily or weekly, and let it run. Tusk pitched the whole product as an AI unit test generator, and this was the version of that pitch with the least setup. Tusk said customers saw around a 30% coverage increase in a month, a vendor figure worth treating as a direction rather than a guarantee.
Self-Healing Test Suites
Existing tests drift out of date as business logic moves. Tusk updated failing suites on each commit so they kept matching current behavior. It's a real convenience and a real risk at once, because a test that rewrites itself to pass can also rewrite itself past a genuine bug.
Context From Jira, Linear, and Your Existing Tests
Tusk pulled business context from Jira and Linear tickets and code patterns from the tests your team already wrote. Generated tests then followed your naming and fixture conventions instead of inventing a new style. Teams that tried generic AI test generators usually cited this as the difference they noticed first.
Code Review and Bug Fixing
The original agent never fully disappeared. Tusk still reviewed pull requests, and it could trace a customer bug report into a fix. Code review came with the same caveat as the tests, in that anything generated needs a human read before merge.
Pros and cons
Pros
- Generated tests arrived already executed, cutting the back and forth that AI copilots usually create.
- Automated test generation fit into CI, so coverage work happened without a separate sprint.
- Integrations with GitHub, Jira, and Linear kept context in the tools engineers already used.
- Pricing per active developer with no seat minimum made it cheap to pilot on one repository.
Cons
- The product is gone. Repositories lost the GitHub App, and Drift users had to remove SDK code from their services.
- Vendor benchmarks were self-reported, so claims about catching edge cases that other agents missed were never independently verified.
- Best results depended on an existing test suite and a single platform. Teams outside GitHub, or with no tests at all, got much less from it.
- AI code review still needs a human pass, and a test that passes isn't proof that the code is correct.
Frequently asked questions
Tusk generated unit and API tests for pull requests, reviewed code, and fixed customer-reported bugs. Engineering teams used it to raise code coverage and catch regressions before deploy.
Related content
Explore related tools, skills, and articles for Tusk.
Tusk Alternatives
Base Code
Base44 (Wix) · Coding · ProductivityBase Code is a shared cloud development environment from Base44. Teammates describe changes to an existing GitHub codebase in chat, preview them live and send them back as pull requests.
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
GitGlow
GitGlow · Coding · ProductivityGitGlow is a local-first desktop Git client that lets independent developers review their own code and their AI agents' changes in one place before shipping.