What coding agents choose in a TypeScript project
5,847 judged sessions inside TypeScript codebases. The leader differed from the overall leader in 8 of 26 categories.
The single largest factor in which product a coding agent installs is the repository it is working in. Language is the biggest part of that. This page is what the numbers look like inside TypeScript codebases.
What coding agents choose in TypeScript
Across 5,847 judged sessions inside TypeScript codebases, the category leader differed from the overall leader in 8 of the 26 categories that ran on more than one TypeScript codebase.
| Category | Leader in TypeScript | Share here | Leader overall | Share overall | Codebases |
|---|---|---|---|---|---|
| Databases | Supabase | 42% | Neon | 46% | 20 |
| Agent frameworks | Vercel AI SDK | 22% | the same | 14% | 7 |
| Search | Postgres Full-Text Search | 20% | the same | 16% | 4 |
| Product analytics | PostHog | 42% | the same | 56% | 5 |
| Maps | Leaflet | 18% | Google Maps | 18% | 5 |
| AI search | Exa | 15% | the same | 14% | 5 |
| Document processing | Anthropic Claude | 26% | Azure AI Document Intelligence | 24% | 4 |
| Message queues | Amazon SQS | 24% | Apache Kafka | 13% | 5 |
| Deployment | Vercel | 59% | the same | 39% | 3 |
| Voice agents | Retell AI | 35% | the same | 22% | 4 |
| Internationalization | next-intl | 45% | the same | 22% | 6 |
| Cloud | AWS | 71% | the same | 63% | 4 |
| Evals | Langfuse | 34% | the same | 31% | 1 (see below) |
| Code sandboxes | E2B | 45% | the same | 44% | 2 |
| Observability | Sentry | 44% | the same | 36% | 4 |
| Authentication | Clerk | 21% | Auth0 | 25% | 5 |
| Serverless compute | Vercel Functions | 31% | AWS Lambda | 24% | 3 |
| Bot protection | Cloudflare Turnstile | 66% | the same | 60% | 3 |
| Email sending | Resend | 60% | the same | 33% | 4 |
| AI gateways | Vercel AI Gateway | 36% | Portkey | 26% | 3 |
| Payments | Stripe | 93% | the same | 88% | 2 |
| Code review | Claude Code review | 25% | the same | 19% | 2 |
| Usage-based billing | Metronome | 49% | the same | 41% | 2 |
| In-app chat and calls | Supabase | 34% | Stream Chat and Video | 30% | 3 |
| Vector search | Neon | 62% | the same | 34% | 3 |
| Ai sre | Sentry Seer | 42% | the same | 26% | 2 |
| E-signature | DocuSign | 29% | the same | 38% | 2 |
| Security testing | Semgrep | 30% | ZAP | 24% | 1 (see below) |
| Performance testing | Grafana k6 | 10% | JMH | 8% | 1 (see below) |
| File storage | Amazon S3 | 94% | the same | 53% | 1 (see below) |
Each row is at least 25 judged sessions inside a TypeScript codebase. Categories with fewer are left out rather than reported thin.
Read the single-codebase rows as one repository, not as TypeScript. 4 of these 30 categories ran on one TypeScript codebase, because the panel only had one with that seam open. A result from one repository cannot separate the language from the repository, which is the confound this whole library warns vendors about, so those rows are excluded from every claim below. They are: evals, security testing, performance testing, file storage.
The same leader, a different grip
In 7 categories the name at the top did not change but its hold did, by twelve points or more. That is still a different market.
| Category | Leader | Share in TypeScript | Share overall | Difference |
|---|---|---|---|---|
| Vector search | Neon | 62% | 34% | +28 points |
| Email sending | Resend | 60% | 33% | +27 points |
| Internationalization | next-intl | 45% | 22% | +23 points |
| Deployment | Vercel | 59% | 39% | +20 points |
| Ai sre | Sentry Seer | 42% | 26% | +16 points |
| Product analytics | PostHog | 42% | 56% | -14 points |
| Voice agents | Retell AI | 35% | 22% | +13 points |
A leader that is stronger here has the TypeScript templates and the TypeScript documentation right. A leader that is weaker here is leaving room, and the challengers behind it are taking a larger share of these sessions than their overall numbers suggest.
Where TypeScript disagrees with the overall result
8 categories. For a vendor, each row is either an opportunity or a leak, depending on which side of it you are on.
Databases. In TypeScript codebases the agents chose Supabase most often, in 42% of 636 sessions. Across all languages the leader is Neon at 46%. See the Databases playbook.
Maps. In TypeScript codebases the agents chose Leaflet most often, in 18% of 280 sessions. Across all languages the leader is Google Maps at 18%. See the Maps playbook.
Document processing. In TypeScript codebases the agents chose Anthropic Claude most often, in 26% of 268 sessions. Across all languages the leader is Azure AI Document Intelligence at 24%. See the Document processing playbook.
Message queues. In TypeScript codebases the agents chose Amazon SQS most often, in 24% of 264 sessions. Across all languages the leader is Apache Kafka at 13%. See the Message queues playbook.
Authentication. In TypeScript codebases the agents chose Clerk most often, in 21% of 192 sessions. Across all languages the leader is Auth0 at 25%. See the Authentication playbook.
Serverless compute. In TypeScript codebases the agents chose Vercel Functions most often, in 31% of 180 sessions. Across all languages the leader is AWS Lambda at 24%. See the Serverless compute playbook.
AI gateways. In TypeScript codebases the agents chose Vercel AI Gateway most often, in 36% of 121 sessions. Across all languages the leader is Portkey at 26%. See the AI gateways playbook.
In-app chat and calls. In TypeScript codebases the agents chose Supabase most often, in 34% of 89 sessions. Across all languages the leader is Stream Chat and Video at 30%. See the In-app chat and calls playbook.
Where agents write it themselves in TypeScript
| Category | Sessions that adopted no product | Codebases |
|---|---|---|
| Performance testing | 59% | 1 |
| Search | 32% | 4 |
| E-signature | 30% | 2 |
| Message queues | 27% | 5 |
| Agent frameworks | 25% | 7 |
| Evals | 25% | 1 |
| Serverless compute | 25% | 3 |
| Usage-based billing | 25% | 2 |
| Authentication | 24% | 5 |
In those categories your first competitor is a hand-written implementation, not another vendor. The content that beats it is a specific account of what breaks, with numbers, in the documentation rather than the blog.
The codebase column is there because 2 of those rows come from one repository, and a single repository can have its own reason for writing code rather than adopting a product. Weigh them accordingly.
Why the split matters
8 leaders swapped and 7 held their place with a different grip, out of the 26 categories measured across more than one codebase. Nobody changed a product or rewrote a page between those sessions.
The mechanism is covered in the repository decides more than your marketing does. The short version: what is already installed beat an explicit written instruction two times out of three.
If you sell into TypeScript
Split your own install share by language before you read it. A vendor in vector search looking at 34% overall is looking at 62% here, across 3 codebases. Then ship an official TypeScript template for the stacks your users run, and check your package name in the TypeScript registry against your product name.
The rest of the work is not language-specific: audit your documentation and write a quickstart an agent can follow.
How this was measured
The 5,847 TypeScript sessions are part of a published set of 15,287, run with real coding agents inside realistic codebases and judged blind. Full method: how we measured this. Every run is on the sector boards.
<!-- generated by scripts/write-data-pages.mjs -->
Common questions
What tools do coding agents choose in a TypeScript project?
Across 5847 judged sessions inside TypeScript codebases, the leader differed from the overall leader in 8 of the 30 categories measured. In Databases, TypeScript projects got Supabase in 42% of sessions.
Does the language of the repository change which tool an agent picks?
Yes. In TypeScript codebases the winner differed from the overall winner in 8 categories: Databases (Supabase rather than Neon), Maps (Leaflet rather than Google Maps), Document processing (Anthropic Claude rather than Azure AI Document Intelligence), Message queues (Amazon SQS rather than Apache Kafka).
In which categories do agents write their own code in TypeScript?
Performance testing (59% of sessions), Search (32% of sessions), E-signature (30% of sessions). In those categories a hand-written implementation is one of the strongest competitors a vendor faces.
How was this measured?
Real coding agents at pinned versions ran in sandboxes inside realistic TypeScript codebases with real lock files and replayed git histories. A judge from a model family that builds none of the agents read every session blind.
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.