Traceback is a software testing suite for engineering, product, and design teams — one place to run tests, catch regressions, and review what changed.
Works across your stack
Product
Computer use
No selectors, no scripts. A virtual cursor and keyboard navigate and interact with your product the same way a person does. The agent sees the rendered screen and decides what to do next.
CI integration
Tests run on every push and post results directly back to the PR. Your team sees pass and fail before anything merges.
3
Tests passed
0
Failed
8.2s
Run time
Visual QA
Compare rendered components against design specs automatically. Drift is caught in review, not after launch.
2 visual differences detectedReliability
When your UI changes, tests adapt. No brittle selectors or broken flows waiting to be noticed.
Auto-healed · 0 broken testsLive viewer
A real-time run viewer shows the agent navigating your product step by step, with a full replay once the run is complete.
Issue tracking
Create Linear issues from failed runs automatically. They close when the test passes again.
LIN-847 opened from run #147How it works
Three steps. Each builds on the previous. No yak-shaving in between.
Install the CLI, wire MCP, or open the web app. Point at anywhere you ship.
Write a prompt or compose a step canvas. Watch it run in a real browser.
Failures land as issues with evidence and a proposed fix. Push to Linear or GitHub.
On the cached-session sign-in path, the Pay button doesn't fire its click handler. Auth state hydration likely races the button mount.
Platforms
Same definitions, same self-healing, same workflow — across every surface your team ships.
Autonomous + deterministic QA for any web app
iOS and Android via Appium
Solana dApp QA with on-chain awareness
Visual QA against Figma
Why Traceback
Three things every team picks Traceback for. None of them require a new framework or a rewrite.
When a class name shifts or a flow gets reordered, Traceback patches the selector, logs the resolution, and keeps the run going. Maintenance load doesn't grow with the suite.
Selector healed
Web flows run in actual browsers. Mobile runs on real iOS and Android via Appium. Web3 runs on mainnet forks with on-chain assertions. No simulation that hides the bugs you actually have.
Trigger runs from your editor through MCP, from a PR, from the CLI, from Slack, or on a schedule. Results post inline on the diff. Issues route into Linear or GitHub.
Triggers
Get started
A short demo is enough to understand how it fits into your workflow.