Modern software runs across increasingly complex, distributed systems, which makes diagnosing and preventing failures one of the hardest problems in engineering. When an incident strikes, on-call engineers must correlate signals from logs, traces, metrics, recent deployments, and infrastructure changes under intense time pressure. Phoebe, a London-based startup founded in 2024, aims to automate that work by building what it calls an immune system for software, powered by swarms of AI agents.
Phoebe's agents continuously investigate live operational data, logs, traces, commits, and infrastructure signals, in a read-only, vendor-agnostic manner that layers on top of a team's existing observability tools rather than replacing them. The agents proactively detect anomalies, form and test hypotheses about root causes, and propose fixes. Crucially, those fixes can take the form of concrete code or configuration changes delivered as pull requests, and in some cases Phoebe can predict and prevent problems before they cause downtime, dramatically shortening or eliminating incidents.
The company reports that its platform can reduce incident resolution times by up to 90 percent and is already trusted by engineering organizations including Trainline, PPRO, Freetrade, Volt, Yapily, and Lindus Health. Its founders bring deep operational pedigree: Matt Henderson and James Summerfield previously built and sold their first startup, Rangespan, to Google, and have backgrounds at Stripe, Google, and Amazon.
Phoebe raised a $17 million seed round in 2025 led by GV (Google Ventures) and Cherry Ventures, with GV partner Roni Hiranand joining the board. With this funding, Phoebe is scaling its agent platform and pursuing a vision in which software systems largely heal themselves, freeing engineers from the relentless burden of firefighting production incidents.