Agent leaderboards / All sectors / Search
Search: which products coding agents choose
Agents wrote search themselves in about 23% of runs.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked three agents to add search to 11 small apps, 459 runs in all, in different wordings and as four different people. Writing it in-house came out on top. Among the named products Postgres Full-Text Search led with about 17%, and OpenSearch came second.
No case held to one answer
We counted 33 cases, each one codebase paired with one agent. In every one, repeated asks in changed wording did not settle on a single product.
The persona changed the order
Asked as an enterprise team, the agents picked OpenSearch 36 times. Asked as senior engineers, the top pick was Postgres Full-Text Search, with Algolia close behind at 23.
Each agent had its own favourite
Cursor picked Meilisearch most, 26 times. Claude Code picked Postgres Full-Text Search 34 times. Codex split almost evenly between OpenSearch and Postgres Full-Text Search.
- Oracle Indexed Search won 23 times, all of them in the Java telecom platform.
- Atlas Search and MongoDB Text Search won only in the Express API service, which runs on Mongoose.
- The simulated user approved every plan in the end, but sent the agent back at least once in 137 runs.
- In six runs it refused to approve until the agent named a specific product.
The ranking459 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | Built in-houseoutcome | 106 | 23% | |
| 2 | Postgres Full-Text Searchpostgresql.org | 79 | 17% | |
| 3 | OpenSearchopensearch.org | 59 | 13% | |
| 4 | Typesensetypesense.org | 42 | 9% | |
| 5 | Meilisearchmeilisearch.com | 40 | 9% | |
| 6 | Oracle Indexed Searchoracle.com | 23 | 5% | |
| 7 | Algoliaalgolia.com | 23 | 5% | |
| 8 | Elasticsearchelastic.co | 22 | 5% | |
| 9 | Atlas Searchmongodb.com | 17 | 4% | |
| 10 | Redis Query Engineredis.io | 9 | 2% | |
| 11 | MySQL FULLTEXTmysql.com | 8 | 2% | |
| 12 | MongoDB Text Searchmongodb.com | 8 | 2% | |
| 13 | Fuse.jsfusejs.io | 7 | 2% | |
| 14 | PostgreSQL pg_trgmpostgresql.org | 5 | 1% | |
| 15 | MiniSearchlucaong.github.io | 2 | 0% | |
| 16 | SQLite FTSsqlite.org | 1 | 0% | |
| 17 | Aurora PostgreSQLaws.amazon.com | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex · GPT-5.6 Sol155 runs | OpenSearch · 27then Postgres Full-Text Search · 26 |
| Cursor · Grok 4.6152 runs | Meilisearch · 26then Postgres Full-Text Search · 19 |
| Claude Code · Claude Opus 5152 runs | Postgres Full-Text Search · 34then Typesense · 18 |
By persona
| Junior developer193 runs | OpenSearch · 23then Meilisearch · 22 |
| Senior engineer106 runs | Postgres Full-Text Search · 28then Algolia · 23 |
| Enterprise team95 runs | OpenSearch · 36then Oracle Indexed Search · 14 |
| Vibe coder65 runs | Postgres Full-Text Search · 23then Fuse.js · 7 |
By what the ask stressed
| The plain ask451 runs | Postgres Full-Text Search · 78then OpenSearch · 57 |
A case is one codebase with one agent, asked several times in different words and as different people. 33 of 33 cases did not hold to a single product.
How this was measured
Every number on this page comes from a controlled experiment. We took 11 small applications, asked 3 coding agents (Codex (GPT-5.6 Sol), Cursor (Grok 4.6), Claude Code (Claude Opus 5)) to add search to each of them, in several wordings and as a junior developer and senior engineer and enterprise team and vibe coder, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 459 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 137 runs. Read the methodology and the publications.
Open the interactive boardThis page as Markdown