Do coding agents recommend OpenRouter?
OpenRouter was chosen in 15% of 329 judged AI gateways sessions, ranking fourth. Measured with Claude Code, Codex, Cursor, Grok Build CLI and Muse Code.
OpenRouter was chosen in 15% of 329 judged AI gateways sessions, ranking fourth. It was also raised as a candidate in 169 further sessions without being chosen.
This page reports what happened when Claude Code, Codex, Cursor, Grok Build CLI and Muse Code had to solve a problem in AI gateways inside a realistic codebase. Not what a chat assistant says about OpenRouter. What an agent actually installed.
The numbers
| Category | AI gateways |
| Sessions in the category | 329 |
| Sessions where OpenRouter was chosen | 49 |
| Install share | 15% |
| Rank in category | 4 of 9 |
| Codebases it won in | 4 |
| Raised as a candidate, not chosen | 169 |
| Chosen when considered | 22% |
| Site | openrouter.ai |
By agent
The agents treat OpenRouter very differently. Cursor chose it in 23% of its sessions in this category and Codex in 6%, a spread of 17 points. Grok Build CLI ran fewer than 30 sessions here, too few to compare.
| Agent | Sessions | Chose OpenRouter | Share |
|---|---|---|---|
| Claude Code | 69 | 8 | 12% |
| Codex | 96 | 6 | 6% |
| Cursor | 71 | 16 | 23% |
| Grok Build CLI | 24 | 5 | 21% |
| Muse Code | 69 | 14 | 20% |
A single blended install share would report the midpoint and hide both ends. If most of your users run Codex, your real position here is 6%, not 15%.
By who was asking
OpenRouter does much better with one kind of buyer than another. It won 38% of sessions asked as vibe coder and 1% of those asked as enterprise team.
| Who is asking | Sessions | Chose OpenRouter | Share |
|---|---|---|---|
| Vibe coder | 40 | 15 | 38% |
| Junior developer | 81 | 24 | 30% |
| Senior engineer | 126 | 9 | 7% |
| Enterprise team | 82 | 1 | 1% |
What OpenRouter was up against
The full ranking in AI gateways, from the same sessions:
| # | Product | Runs won | Share |
|---|---|---|---|
| 1 | Portkey | 86 | 26% |
| 2 | Cloudflare AI Gateway | 76 | 23% |
| 3 | Vercel AI Gateway | 52 | 16% |
| 4 | OpenRouter (this page) | 49 | 15% |
| 5 | LiteLLM | 25 | 8% |
| 6 | Built in-house (no product adopted) | 15 | 5% |
| 7 | Amazon Bedrock | 14 | 4% |
| 8 | Helicone | 1 | 0% |
What this means
A solid second or third position in a category means the agent is genuinely choosing rather than reaching automatically. That is a winnable position, because the inputs it uses can be changed.
The fastest gains are usually in the sessions that were nearly won: read them, find the step where the agent moved on, and fix that step.
Where these numbers come from
The 329 sessions in AI gateways are part of a published set of 15,287, run with real coding agents inside realistic codebases and judged blind. The full method is on one page: how we measured this.
Every AI gateways run can be replayed on the board.
If you work on OpenRouter: the judge recorded a reason for every session where it was raised and passed over. Those reasons are in the transcripts.
<!-- generated by scripts/write-data-pages.mjs -->
Common questions
Do coding agents recommend OpenRouter?
Yes. OpenRouter was chosen in 49 of the 329 judged sessions in AI gateways, a 15% install share, ranking fourth in its category.
Does Claude Code recommend OpenRouter?
In 8 of the 69 sessions in AI gateways run with Claude Code, which is 12%.
Do different coding agents treat OpenRouter differently?
Yes, and by a wide margin. Cursor chose it in 23% of its runs and Codex in 6%.
How was this measured?
Real coding agents at pinned versions were run in sandboxes inside 110 realistic codebases and asked to solve real tasks. A simulated project owner approved or questioned each recommendation before any code was written, and a judge from a model family that builds none of the agents read every session blind.
How often is OpenRouter considered but not chosen?
It was raised as a candidate in 169 sessions without being chosen, and chosen in 49. That is a 22% conversion from considered to chosen.
Where this comes from
Armature ran 15,287 judged sessions with Claude Code, Codex, Cursor, Grok Build CLI and Muse Code inside 110 realistic codebases, and published every run. The numbers on this page come from that work.