Resolve AI raises $40M extension at $1.5B valuation to build agents that debug production systems

Resolve AI just raised a $40M extension round at a $1.5B valuation, two months after closing a Series A at $1B. The company, founded just over a year ago, builds AI agents that debug and operate production software systems — the operational counterpart to coding agents that generate code.

Spiros Xanthos frames the problem simply: as software production accelerates, particularly with AI-generated code, the attack surface for production failures grows faster than human on-call teams can handle. Resolve sits as the first line of defense, catching production issues before they reach end users and generating resolutions to compress the response loop. Customers include Coinbase, Salesforce, and MongoDB.

“We are building agents that can help you debug and run production — the counterpart to coding agents that produce all this code. Look at some of our customers: Coinbase, Salesforce, MongoDB. We just did the Series A at a billion dollars like two months ago and now we've announced a $40M extension at $1.5B.”

What the product does

Unlike agents that operate on a code repository in isolation, Resolve works across the full production stack — code, telemetry, logs, metrics from tools like Datadog and Splunk, and cloud infrastructure like AWS. Xanthos describes it as a "production IDE," analogous to what a developer environment does for writing code.

The harder technical problem is training agents to handle the long, multi-step diagnostic chains that production debugging requires — iterating across code, infrastructure, and observability tooling over many steps. Xanthos argues that kind of sequential planning behavior is largely absent from existing foundation model training sets, which is why Resolve is now building its own domain-specific models. The company announced a dedicated research lab alongside the funding.

On data, the training signal isn't customer codebases. It's the actions humans take when solving production problems — sequences that, Xanthos argues, general-purpose models haven't been trained to replicate. He's direct that the big labs have overstated how hard it is for domain-focused companies to compete on model quality.

Growth and valuation

Demand has been inbound since launch, Xanthos says, driven by the fact that production reliability is a well-understood and high-stakes problem for any company delivering through software. Resolve focuses primarily on large enterprises, where system complexity makes the problem acute. The go-to-market combines product-led and sales-led motions.

On the round structure, Xanthos is candid: low dilution follows from being in a strong position, not from negotiating tactics. His prior startup experience taught him that fundraising is a lagging indicator, not a milestone.

The $500M valuation step-up in two months, with no lead investor named in the conversation, points to a competitive dynamic among investors — though Xanthos doesn't elaborate on terms beyond the headline figures.