Do coding agents recommend OpenSearch?
OpenSearch was chosen in 13% of 459 judged search sessions, ranking second. Measured with Claude Code, Codex and Cursor.
OpenSearch was chosen in 13% of 459 judged search sessions, ranking second. It was also raised as a candidate in 231 further sessions without being chosen.
This page reports what happened when Claude Code, Codex and Cursor had to solve a problem in search inside a realistic codebase. Not what a chat assistant says about OpenSearch. What an agent actually installed.
The numbers
| Category | Search |
| Sessions in the category | 459 |
| Sessions where OpenSearch was chosen | 59 |
| Install share | 13% |
| Rank in category | 2 of 16 |
| Codebases it won in | 3 |
| Raised as a candidate, not chosen | 231 |
| Chosen when considered | 20% |
| Site | opensearch.org |
By agent
The three agents land within 7 points of each other on OpenSearch, which is closer than most products in this experiment manage.
| Agent | Sessions | Chose OpenSearch | Share |
|---|---|---|---|
| Claude Code | 152 | 17 | 11% |
| Codex | 155 | 27 | 17% |
| Cursor | 152 | 15 | 10% |
By who was asking
OpenSearch does much better with one kind of buyer than another. It won 38% of sessions asked as enterprise team and 0% of those asked as senior engineer.
| Who is asking | Sessions | Chose OpenSearch | Share |
|---|---|---|---|
| Vibe coder | 65 | 0 | 0% |
| Junior developer | 193 | 23 | 12% |
| Senior engineer | 106 | 0 | 0% |
| Enterprise team | 95 | 36 | 38% |
What OpenSearch was up against
The full ranking in search, from the same sessions:
| # | Product | Runs won | Share |
|---|---|---|---|
| 1 | Built in-house (no product adopted) | 106 | 23% |
| 2 | Postgres Full-Text Search | 79 | 17% |
| 3 | OpenSearch (this page) | 59 | 13% |
| 4 | Typesense | 42 | 9% |
| 5 | Meilisearch | 40 | 9% |
| 6 | Oracle Indexed Search | 23 | 5% |
| 7 | Algolia | 23 | 5% |
| 8 | Elasticsearch | 22 | 5% |
What this means
This is an integration problem, not a presence problem. Agents raised OpenSearch in 231 sessions and chose it in 59, so it reaches the shortlist and then loses. Something at the last step is costing the session, and in our data that is usually a quickstart that does not run when pasted, documentation describing an interface that changed, or a package name that does not match the product name.
That is the cheaper of the two problems to have. The reason is written down in each losing transcript.
Where these numbers come from
The 459 sessions in search are part of a published set of 5,292, run with real coding agents inside realistic codebases and judged blind. The full method is on one page: how we measured this.
Every search run can be replayed on the board.
If you work on OpenSearch: 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 OpenSearch?
Yes. OpenSearch was chosen in 59 of the 459 judged sessions in search, a 13% install share, ranking second in its category.
Does Claude Code recommend OpenSearch?
In 17 of the 152 sessions in search run with Claude Code, which is 11%.
Do different coding agents treat OpenSearch differently?
Not much. The three agents chose it at similar rates, between 10% and 17% of their runs.
How was this measured?
Real coding agents at pinned versions were run in sandboxes inside 51 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 OpenSearch considered but not chosen?
It was raised as a candidate in 231 sessions without being chosen, and chosen in 59. That is a 20% conversion from considered to chosen.
Where this comes from
Armature ran 5,292 judged sessions with Claude Code, Codex and Cursor inside 51 realistic codebases, and published every run. The numbers on this page come from that work.