Open data

BYOAgent

Every byte your project owns is exportable. We're a primitive — your agent brings its own AI; we provide the typed surface and the storage. The competition isn't Typeform or Qualtrics — it's the agent that thinks “I'll just spin up Postgres and a form on Vercel.” We win when the cost of using us is less than the cost of that.

What we don't run

No LLM calls on our infrastructure. Ever.
No theme or sentiment analysis on response content.
No AI-generated surveys from prose.
No auto-translation, auto-grading, or auto-anything that costs us tokens.

Your agent does that work, in your context, on your token budget. We surface the raw shapes it needs.

Take your data — every path

One bearer key (service key or OAuth access token) gets you to all of it. Available via REST or MCP — same data either way.

GET /api/v1/projects/[projectId]/export

Full project bundle: surveys, versions, responses, members, audit events. ?format=jsonl streams as newline-delimited JSON.

GET /api/v1/surveys/[surveyId]/responses?format=csv

Raw responses as CSV, columns derived from the survey's Question[] schema.

GET /api/v1/surveys/[surveyId]/questions/[questionId]/samples

Raw values for one question — designed for agent-side clustering / sentiment / topic modeling. Up to 10k samples per call.

GET /api/v1/projects/[projectId]/activity?format=jsonl

Full audit log of every action on the project — responses, edits, publishes, agent calls, OAuth grants — as JSONL.

GET /api/v1/surveys/[surveyId]/summary

Token-efficient analytics — buckets, NPS, free-text clusters (keyword frequency, intentionally simple — bring your own clusterer for better).

MCP equivalents

The same surface via MCP, for agents:

exportProject({ projectId, includeAgent? }) — full bundle in one call
queryResponses — raw paginated rows
samplesByQuestion — per-question raw values
summarizeResponses — bucketed analytics, prompt-ready
exportInsights — headline + bullets + summary

If you want to leave

Call GET /api/v1/projects/[projectId]/export?format=jsonl and save the file.
That file contains every survey, every version, every response, every member, every audit event. Schema is documented at /docs/schema and /openapi.json.
Delete the project from the dashboard. After 30 days the data is permanently removed from our systems (per the Terms of service).

That's it. No support ticket, no “data retrieval fee,” no proprietary format.

Why this is the policy

We're not VC-backed. We win on time-cost, not on lock-in. An agent that knows it can take its data out at any moment is more likely to start with us; that calculus only works if the export path is real, fast, and complete.

A survey tool that holds your responses hostage is competing against the agent rolling its own. We'd rather be the obvious-cheaper-than-rolling-your-own option, every time.

Machine-readable manifest at /.well-known/survey. OpenAPI spec at /openapi.json.