Traversal, a New York company that sells AI agents for site reliability work, announced a $120 million Series B at a $1 billion valuation. The round was co-led by Kleiner Perkins, Redpoint Ventures and Khosla Ventures, with significant participation from Sequoia and Battery Ventures. Kleiner Perkins and Sequoia both increased earlier stakes, according to the company. Deal counsel Goodwin confirmed the size, valuation and lead investors.
The Next Web reports total funding of $173 million. Traversal came out of stealth in June 2025 with $48 million from Sequoia and Kleiner Perkins, and Amex Ventures invested in December 2025, per that account. The company says annual recurring revenue has grown nine times this year; chief executive Anish Agarwal declined to give Forbes an absolute revenue figure, according to TNW.
What the product claims to do
Site reliability engineering is the work of keeping large software systems running. When something fails, engineers dig through logs, metrics, deployments and code to find the cause. Traversal says its agents run that search with a Production World Model that maps telemetry for AI, and a Causal Search Engine that tests many causes in parallel. The company formally introduced those pieces in March 2026.
That is a different bet from coding agents that write software. We have covered what an AI agent is and where it fails. We have also looked at tools that give coding agents full access to a developer’s machine or let them trigger cloud workflows without rewriting them. Traversal sells the opposite end of the loop: agents that investigate production once new code is live.
On its blog, Traversal cites survey figures it did not independently attribute. It says 42% of developers now use AI for a majority of their code, up from 12%, and that 52% spend more time debugging in production. It also cites unplanned outages costing Global 2000 companies $600 billion in 2026, up 50% in two years. Treat those as company-cited figures, not independently verified.
Alex Palcuie, SRE Lead at Anthropic, is quoted on the Traversal blog. “We can’t trust LLMs for incident response… The problem is its inability to step back and start discerning between causation and correlation… For us humans, it is hard as well.” Traversal’s pitch is that causal search, not plain correlation from a language model, makes incident response reliable enough to trust.
Growth claims and who uses it
Traversal says monthly investigations rose from 513 to nearly 73,000 this year. It says about two-thirds of usage now goes beyond root cause analysis into alert triage, release risk, CMDB validation, security, customer journey health and production support. It also ships Traversal Workers (proactive agents that start when an alert fires) and Traversal MCP (a feed of production intel into coding agents).
Customers named by the company, as reported by TNW, include Nubank, Harvey, DigitalOcean, PepsiCo and American Express. Nubank founder and chief executive David Vélez said in that report: “Five weeks into our Traversal rollout, and it’s already becoming part of how our on-call engineers work incidents.”
Traversal reports an ORCA-bench suite with Cornell Tech and Columbia University covering 1,079 root-cause tasks. The best of five frontier agents scored 25.3% on medium difficulty, 10% when reports were vague, and hallucination rates up to 40%, the company said via TNW. We have not run the bench. Cofounders Anish Agarwal, Raaz Dwivedi, Raj Agrawal and Ahmed Lone come from causal ML and RL research at MIT, Berkeley, Columbia and Cornell, per the company.
What to do this week
If you buy coding agents and still burn engineer time in war rooms, ask your SRE lead whether root-cause work is measured at all. Then read Traversal’s own announcement against the ORCA-bench numbers it cites. The gap between a language model story and a verified cause is the product claim you are paying for. Ask any vendor for a timed pilot on one recurring incident class, with your own pass/fail criteria, before treating a billion-dollar valuation as proof the outer loop is closed.
Sources
Get the weekly email
The week's most useful AI news, a tutorial and an interview.
Martin Scott
Martin Scott is the editor for business and policy at AI Observer, covering startups and deals, regulation, health and science, and AI agency profiles. He follows the money and the legal text, and separates what a rule says from what people claim it says. His style is analytical and plain-spoken.
Spotted a mistake? Tell us on the corrections page.