The evidence base,
over HTTP.
Your code can read the data that the app reads: studies, transcripts with timecodes, study reports, clips and uploads. All data is plain JSON. Claude Code and Cursor can read the same data through the MCP server.
You create keys in Settings, and you see each key one time. You can revoke a key at any time.
The same endpoints that the app uses
The base URL is https://api.userevaluation.com, and each path starts with /api/. Send a bearer token in the header. Requests and responses are JSON.
Read. List studies and sessions. Get a transcript with speakers and timecodes. Get a study report with its themes, claims and clips.
Write. Create and launch a study. Upload a recording or a document. Make clips, generate a study report and ask Eva a question.
Listen. Add a webhook endpoint for events such as session.completed, readout.ready and file.ready. Then you do not need to poll.
- Studies
GET POST /api/studies · GET PATCH /api/studies/:id · GET /api/studies/:id/estimate · POST /api/studies/:id/publish - Sessions
GET /api/studies/:id/sessions · GET /api/studies/:id/sessions/:sessionId · GET /api/studies/:id/participants - Study reports
GET /api/studies/:studyId/readout · POST /api/studies/:studyId/readout/generate · POST /api/studies/:studyId/readout/exports - Clips and reports
GET POST /api/clips · GET POST /api/reports · POST /api/reports/:id/exports - Library
GET /api/library/files · POST /api/library/uploads · POST /api/library/files/:id/complete - Eva and usage
POST /api/eva/ask · GET /api/search · GET /api/billing/usage
The help centre lists all endpoints, the pagination and the error codes. Read the API guide.
curl "https://api.userevaluation.com/api/studies?limit=20" \
-H "Authorization: Bearer ue_live_..."
# 200 OK
# {
# "items": [
# {
# "id": "...",
# "title": "Onboarding · why trials stall",
# "status": "ready",
# "createdAt": "2026-09-03T10:12:41.000Z"
# }
# ],
# "nextCursor": null
# }curl -X POST https://api.userevaluation.com/api/library/uploads \
-H "Authorization: Bearer ue_live_..." \
-H "Content-Type: application/json" \
-d '{ "name": "call.mp4", "sizeBytes": 52428800, "contentType": "video/mp4" }'
# → The URL for the PUT request, and the file id
curl -X POST https://api.userevaluation.com/api/library/files/{fileId}/complete \
-H "Authorization: Bearer ue_live_..."
# → Processing starts. Then the file.ready webhook arrives.Keys, limits and webhooks
These parts of an API are easy to get wrong. Ours work as you expect, so you can start on your integration.
Bearer keys, stored as a hash
Keys start with ue_live_, and you create them in Settings. We keep only a SHA-256 hash, so we cannot show a key again. A key can be read-only, or it can read and write.
One rate limit
Each key can send 600 requests a minute. Above that limit, you get a 429 with the code rate_limited. Wait, then send the request again.
Signed webhooks
Each delivery has a UE-Signature header with an HMAC-SHA256 signature of the body. Reply with a 2xx status in 10 seconds. If you do not, we try again, up to 5 attempts in total.
Each error has an HTTP status and a JSON body with a code, a message and a requestId. Send the requestId when you contact us.
Your evidence in ChatGPT, Claude and Cursor
The MCP server lets ChatGPT, Claude, Claude Code, Cursor and VS Code read your studies, transcripts and study reports, search what participants said and ask Eva.
One address. The server is at https://api.userevaluation.com/mcp. It uses Streamable HTTP, so there is nothing to install.
Sign in with OAuth. ChatGPT and Claude custom connectors, and Claude Code, sign in with your account. You select the workspace and read or write access, then click Allow. The server supports OAuth 2.1 with PKCE and dynamic client registration.
Or use an API key. Send the key in the Authorization header, as for the API. A read-only key can read and ask Eva.
Same rules as the API. A connection acts as the person who made it. MCP calls count as API calls. Each key or connection can send 600 requests a minute.
Name: User Evaluation
URL: https://api.userevaluation.com/mcp{
"mcpServers": {
"userevaluation": {
"url": "https://api.userevaluation.com/mcp",
"headers": { "Authorization": "Bearer ue_live_…" }
}
}
}Studies
List your studies. Read the question, guide, screener and audience of a study. See its status and progress.
Sessions
List the sessions of a study. Read one session with its full transcript, chapters and signals.
Study reports and reports
Read the themes, claims and clips of a study report. List reports and read one report.
Swarm
List Swarm runs and their findings. Read one finding with its evidence.
Search
Find studies, reports, files and people by name. Search what participants said in transcripts, clips and study report claims.
Library
List your Library files. Read one file with its transcript.
The server also has tools to set up, launch and run studies, invite people, and make clips, reels and reports. Anything that costs money or emails people needs your yes first.
No SDK. Plain HTTP and JSON.
We do not have SDKs or an OpenAPI file. Each endpoint uses plain HTTP and JSON, so any HTTP client works.
curl -H "Authorization: Bearer $UE_API_KEY" https://api.userevaluation.com/api/studies
Use curl to test a new key. A key that is not valid gets a 401.
await fetch(url, { headers: { Authorization: `Bearer ${key}` } })
Use the built-in fetch. When a request fails, read error.code from the JSON body.
requests.get(url, headers={"Authorization": f"Bearer {key}"})
Use requests or httpx. Each list endpoint returns items and nextCursor.
Zapier, Make and n8n can call the endpoints with an API key. They can also receive our webhooks.
Questions about the API
For other questions, send an email to support@userevaluation.com. An engineer replies.
The API is part of the Pro, Team and Scale plans. It is not on the Free plan. On each plan, a key can send 600 requests a minute. Studies that you launch with the API cost the same as in the app.
No. Use plain HTTP and JSON from any language. The help centre lists the endpoints, the pagination and the error codes.
A key acts as the person who created it, with their workspace role, up to Member rights. A read-only key can send GET requests and ask Eva. Other requests get a 403.
Yes. A launch charges your saved card, as in the app. First, get the cost with GET /api/studies/:id/estimate. Send the total as expectedTotalCents. If the price changes, you get a 409 and we charge nothing.
You get a 429 with the code rate_limited. Wait, then send the request again. To know when a job is done, use webhooks. Do not poll.
If you do not want to write code
Integrations do many of the tasks that people build with the API. Eva answers questions from your evidence without code.
Integrations
Import recordings from Zoom, Google Drive and Dropbox. Send events to Slack, Notion and Linear with webhooks and Zapier.
See all integrations →The evidence baseAnalysis
Interviews, agent sessions, uploads and call recordings are in one evidence base. Each answer cites a clip.
See how it works →Get a key. Build the integration.
Create a key in Settings and send one curl request. Then your code can read your evidence base.
The API is part of Pro, Team and Scale. Each key can send 600 requests a minute.