Helping coding agents solve problems on the most complex codebases
Undo gives agents the runtime context they need to solve the most complex problems
AGENTS NEED CONTEXT TO WORK ON COMPLEX CODE
Mission-critical codebases are large, complex and unstructured. The hardest problems live in runtime state — execution paths, values flowing across module boundaries, effects far from their cause. Source code doesn’t capture it. Logs don’t capture it. So agents fill the gap with plausible guesses, and plausible isn’t good enough.
PURPOSE-BUILT FOR THE CODE THAT MATTERS MOST
Undo closes the gap. We produce a deterministic recording of runtime execution: every state change, every call, across every module. Agents query the recording to reason about what the program actually did — the way they already reason about what the code says. The result is fully automated root cause analysis, on problems that once took months of engineering effort or required your best engineers.
That’s why thousands of developers rely on Undo to prevent outages, keep codebases maintainable, and get more from the agents they already use.
WHAT PRODUCTS ARE AVAILABLE?
Undo started out as a fast-growing technology start-up founded by Greg Law and Julian Smith. Both began working evenings and weekends in Greg’s garden shed, determined to build a tool that would ease the pain of working with complex software.
For over a decade, Undo established itself as the time travel debugging company for Linux. In 2026, software engineering changed: AI coding agents took on work that had always belonged to humans, and the limits of what they could reason about became the new bottleneck. Undo’s technology now gives those agents the runtime context they need to solve the most complex problems on the most complex codebases.
LATEST NEWS
Coding agents need context
AI agents don't have an intelligence problem. They have a context problem.
Undo hosts 2026 Customer Council
In March, we ran our 2026 Customer Council meeting in California, hosting senior and principal engineering leads from the world's largest technology organisations.
Shaping the future of AI debugging
AI code assistants can do a lot; but when code breaks, they're just guessing. Undo makes AI coding assistants better by giving it actionable context.