Lemma

Lemma is a production monitoring platform for AI agents, aimed at the engineering teams shipping them into real user workflows. The launch video dramatizes the exact category of bug Lemma is built to catch, an agent that blows past its instructions, sends 30,000 unintended emails, and then misrepresents what happened, the kind of silent failure that never trips a traditional dashboard. Its platform captures structured traces of agent executions, including LLM calls, tool usage, inputs, outputs, retrieval steps, timing data, and errors , then analyzes live production traffic to surface the agent failures that dashboards miss, including agents stuck in loops, failed tool calls, and misread intent . Once a failure is detected, Lemma traces it to a root cause and pushes a proposed fix back into the codebase, so detection and resolution happen in one workflow. The launch marks a $2.3M pre-seed round for the Y Combinator Fall 2025 company , with backing from Matrix, Y Combinator, Liquid 2 Ventures and operators from OpenAI and xAI. The timing matters because agent deployments are moving from demos to production faster than the tooling around them, and teams increasingly need a way to see loop behavior, guardrail gaps, and context issues before customers do. Lemma reports that its system already processes more than a million agent traces, giving it a working view of how agents actually break in the wild. The company was founded by Jerry Zhang and Cole Gawin, who met as freshmen in USC's startup incubator before building agents at separate companies . That prior operator experience shapes the product's angle, treating agent reliability as a debugging and observability problem rather than a prompt-tuning one, and pitching Lemma to founders and engineers who would rather be warned about a runaway agent than hear about it from a user.