Tusk 🐘 - AI Agent for Generating Unit/Integration Tests

Hey everyone, we’re Sohil and Marcel, the founders of Tusk (YC W24).

Tusk is an AI agent that generates unit and integration tests for your PRs.

We’re the team that built the first publicly available AI coding agent in the market. Now we’re focused on taking on the tedious software engineering work that developers dread the most — writing automated tests.

https://youtu.be/OK_uhIGEQ0Y?feature=shared

Problem

It goes without saying that writing tests is important for maintaining high code quality in any engineering organization.

However, any software engineer who has had to hastily meet product deadlines knows that writing unit and integration tests is the first thing to go. Their main priority in this case is to ship the feature as fast as possible.

The problem is that the lack of tests then results in customers complaining about bugs in the feature. And all too often the blame is assigned to engineering leadership for not caring enough about quality.

We know how painful this experience is from our previous jobs. We’ve experienced instances of forgone tests leading to embarrassing bugs, including one that ruined a live demo for a high-value pilot customer (who you can guess didn’t convert).

Solution

We built Tusk to provide EMs and senior engineers in charge of quality efforts a way to ship fast while improving code quality. Our AI agent generates unit and integration tests using your codebase context and business logic.

Tusk functions as a non-blocking PR check that suggests happy path and edge case tests that have not been covered by the existing test suite. Our agent shows you which tests have passed or failed and allows you to incorporate the test cases into the PR with one click.

Engineering teams at companies like DeepLearning.AI, Wyndly, and Afterword use Tusk to prevent bugs while increasing test coverage.

Tusk is the first AI tool that noticeably accelerated our dev cycle. By scaling our testing infra, it's given us confidence to ship faster and been instrumental in helping our lean team scale across 3 stacks. Of all the AI tools being built right now, I'm most excited about Tusk.

– Zack Moy, Co-Founder & CTO at Afterword

Edge cases that are not covered can cause the most critical issues. By using Tusk, our team gains an additional guardian to protect our product from the threats of edge cases.

– Amo Chen, Senior Backend Engineer at DeepLearning.AI

Our priority as a fast-shipping product team is making sure we get high quality code out the door quickly. Tusk has been instrumental in helping our engineering team increase test coverage with minimal effort.

– Aakash Shah, Founder & CEO at Wyndly (YC W21)

Why Use Tusk

Tusk surpasses other test generation and code review tools in these ways:

We run the tests we generate. Tusk runs the tests it generates in an ephemeral sandbox and iterates on its output so you can be confident that its tests are checking for relevant happy path and edge cases. Code review tools do not do this, which results in vague comments that are hard to trust.

Optimized for test generation quality. 69% of Tusk-generated test suites are incorporated into customers’ PRs. We use more compute to reason through test scenarios. AI copilots in your IDE are optimized for latency as opposed to quality when generating tests.

Developer-centric, fully customizable. Tusk puts engineers in control by allowing them to accept/decline the generated tests and customize configs to meet your team's testing guidelines (how to mock, using factories, avoiding specific directories, etc.).

Read our documentation to learn more.

Interested?

Tusk is now available for growth-stage and enterprise companies looking to auto-generate tests for their PRs and/or backfill tests for their monolith.

Apply for access to see how Tusk can help your team ship faster but safer.