Agent leaderboards / All sectors
Which tools do coding agents choose?
We asked three coding agents, Cursor, Claude Code and Codex, to add one capability to a small app and pick the product themselves. That ran 5292 times, across 18 sectors and 51 apps. Most sectors have a favorite, few have a default.
Key learnings
Five sectors have one obvious winner
In five sectors the leading product took at least half the runs. Stripe is the strongest, with about 88% in Payments. Neon took about 66% in Databases, and AWS about 62% in Cloud.
Half the sectors are still contested
Nine sectors have a leader with between 22% and half the runs. Vercel leads Deploy with about 41%. In AI gateway no product got past 22%.
Sometimes the agents write it themselves
Across every run, the agents built the capability in-house about 13% of the time. In three sectors they did that more often than they picked any product. In Performance in CI it was about 52% of runs.
- In AI gateway the top two split almost evenly, Portkey with 30 wins and Cloudflare AI Gateway with 30.
- In Search each agent led with a different product, and the top one, Postgres Full-Text Search, had about 17%.
- PostHog took about 53% of Product analytics, and the agents wrote their own in about 23% of runs.
- Auth covered 17 apps, more than any other sector, and WorkOS AuthKit led with about 26%.
The sectorsone page each
Agent sandboxes
E2B won about 42% of runs, Modal about 25%.
299 runs · E2B 42%Observability
Sentry won about 37% of runs and led all three agents.
360 runs · Sentry 37%Payments
Stripe won about 88% of the payment runs across 11 small apps.
395 runs · Stripe 88%Deploy
Vercel took about 41% of deploy picks, Render about 34%.
270 runs · Vercel 41%Auth
WorkOS AuthKit led with about 26% and no provider stood out.
201 runs · WorkOS AuthKit 26%Email providers
Resend won about 36% of the runs, Postmark about 27%.
208 runs · Resend 36%Product analytics
PostHog won about 53% of runs across three agents.
359 runs · PostHog 53%Databases
Neon took about 66% of 356 runs picking a database.
356 runs · Neon 66%File storage
Amazon S3 took about 46% of runs, and two products tied behind it.
90 runs · Amazon S3 46%LLM evals & observability
Langfuse won about 34% of runs, with in-house code second.
288 runs · Langfuse 34%Voice Agents
Vapi led with about 24%, but each agent had a different favorite.
308 runs · Vapi 24%Serverless functions
AWS Lambda and Vercel Functions finish close, about 24% and 23%.
287 runs · AWS Lambda 24%Cloud
AWS won about 62% of the runs across six small apps.
215 runs · AWS 62%AI gateway
Portkey and Cloudflare AI Gateway tie on top at about 21% each.
140 runs · Portkey 21%Bot protection
Cloudflare Turnstile won about 57% of 160 runs.
160 runs · Cloudflare Turnstile 57%Search
Agents wrote search themselves in about 23% of runs.
459 runs · Postgres Full-Text Search 17%Agent frameworks
Agents wrote it themselves in about 25% of runs.
681 runs · Vercel AI SDK 14%Performance in CI
Agents wrote the gate themselves in about 52% of runs.
216 runs · JMH 10%