Do coding agents recommend Mapbox?
Mapbox was chosen in 8% of 319 judged maps sessions, ranking fourth. Measured with Claude Code, Codex and Cursor.
Mapbox was chosen in 8% of 319 judged maps sessions, ranking fourth. It was also raised as a candidate in 255 further sessions without being chosen.
This page reports what happened when Claude Code, Codex and Cursor had to solve a problem in maps inside a realistic codebase. Not what a chat assistant says about Mapbox. What an agent actually installed.
The numbers
| Category | Maps |
| Sessions in the category | 319 |
| Sessions where Mapbox was chosen | 25 |
| Install share | 8% |
| Rank in category | 4 of 32 |
| Codebases it won in | 7 |
| Raised as a candidate, not chosen | 255 |
| Chosen when considered | 9% |
| Site | mapbox.com |
By agent
The three agents land within 10 points of each other on Mapbox, which is closer than most products in this experiment manage.
| Agent | Sessions | Chose Mapbox | Share |
|---|---|---|---|
| Claude Code | 108 | 4 | 4% |
| Codex | 111 | 15 | 14% |
| Cursor | 100 | 6 | 6% |
By who was asking
Mapbox does much better with one kind of buyer than another. It won 21% of sessions asked as senior engineer and 1% of those asked as junior developer.
| Who is asking | Sessions | Chose Mapbox | Share |
|---|---|---|---|
| Vibe coder | 64 | 2 | 3% |
| Junior developer | 147 | 2 | 1% |
| Senior engineer | 72 | 15 | 21% |
| Enterprise team | 36 | 6 | 17% |
What Mapbox was up against
The full ranking in maps, from the same sessions:
| # | Product | Runs won | Share |
|---|---|---|---|
| 1 | Google Maps | 52 | 16% |
| 2 | Leaflet | 52 | 16% |
| 3 | Apple Maps (MapKit) | 46 | 14% |
| 4 | Mapbox (this page) | 25 | 8% |
| 5 | MapLibre | 21 | 7% |
| 6 | react-native-maps | 19 | 6% |
| 7 | Azure Maps | 18 | 6% |
| 8 | Built in-house (no product adopted) | 16 | 5% |
What this means
This is an integration problem, not a presence problem. Agents raised Mapbox in 255 sessions and chose it in 25, 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 319 sessions in maps are part of a published set of 6,458, run with real coding agents inside realistic codebases and judged blind. The full method is on one page: how we measured this.
Every maps run can be replayed on the board.
If you work on Mapbox: 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 Mapbox?
Yes. Mapbox was chosen in 25 of the 319 judged sessions in maps, a 8% install share, ranking fourth in its category.
Does Claude Code recommend Mapbox?
In 4 of the 108 sessions in maps run with Claude Code, which is 4%.
Do different coding agents treat Mapbox differently?
Yes, and by a wide margin. Codex chose it in 14% of its runs and Claude Code in 4%.
How was this measured?
Real coding agents at pinned versions were run in sandboxes inside 63 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 Mapbox considered but not chosen?
It was raised as a candidate in 255 sessions without being chosen, and chosen in 25. That is a 9% conversion from considered to chosen.
Where this comes from
Armature ran 6,458 judged sessions with Claude Code, Codex and Cursor inside 63 realistic codebases, and published every run. The numbers on this page come from that work.