YC's Paxel turns AI-coding sessions into a Startup School signal
The free Docker tool reads Claude Code, Codex CLI and Cursor transcripts, generates a builder profile, and tells applicants a Paxel token can boost their chances.
By Staff Writer · Published
Why it matters
Y Combinator is turning AI coding behavior into an application signal, suggesting accelerators may increasingly judge founders by how they build with tools, not just what they claim to have built.

Y Combinator (@ycombinator) launched Paxel, a free tool that reads local Claude Code, Codex CLI and Cursor sessions and turns them into a profile of how founders build with AI, the accelerator said Friday in a two-post thread on X.
https://x.com/ycombinator/status/2062982928574136449
YC says 32,200 sessions have been uploaded and analyzed so far. The product asks users to sign in, run a curl command from one repo or a parent folder of multiple repos, then wait about 15 to 30 minutes for a report. Paxel profiles work across five dimensions: steering, execution, engineering, product instinct and planning. It also assigns archetypes such as Architect, Quality Guardian, Velocity Machine and Night Owl, and surfaces decision patterns and a suggested "growth edge" from the user's actual AI-coding transcripts.
The recruiting hook is explicit. In its FAQ, YC says linking a Paxel token "boosts your chances of getting into Startup School 2026" and directs applicants, including people already on the waitlist, to attach the token to their application. That makes Paxel less a standalone developer utility than a new admissions instrument for an accelerator trying to understand how founders actually use AI coding tools.
YC is also trying to blunt the obvious privacy objection. The accelerator says the analysis runs locally in Docker, and that a user's working tree, .env files and raw transcripts stay on their machine. During analysis, YC says short transcript excerpts, including prompts, agent responses and snippets of tool calls, go to Claude or GPT through its proxy so the system can summarize and score sessions. The final upload to YC is a JSON payload of scores, narratives, redacted decisions and session metadata.

The unresolved questions are the ones that matter for applicants. YC's post and product page do not say how heavily Paxel data will be weighted, what exactly the profile scores reward, or whether the tool will remain tied to Startup School after this application cycle.