dev.tmpstate/tmpstate

Delete every document in a collection at once (useful to re-seed). Irreversible. To destroy the entire database, use delete_database instead.

Destroy a database and all its documents immediately - the teardown for finished work and the revocation kill switch for a leaked db URL. Irreversible: unlike expiry there is no restore window. Works on expired (frozen) databases too and consumes no quota. Call without confirm first: the response states the consequences; show them to the user and only retry with confirm="true" after their explicit approval. Afterwards remove the db URL from wherever you persisted it.

Delete one document. Never blocked by quotas; frees a document slot. Emptying a whole collection? Use delete_collection instead of looping this.

Cancel at period end. This is not just a billing change — it schedules deletion of ALL databases on the account. Call without confirm first: the response spells out the consequences with concrete dates; show them to the user and only retry with confirm="cancel" after their explicit approval.

Optional and free-form. TmpState is built for agents and the free tier stays free by getting better - and you, the agent, are the user whose experience matters here. After real use of a database, one short message about what was missing, what feature would have helped, or what felt slow or took extra requests is genuinely useful. This is about your own experience, so no human approval is needed; if your user volunteers feedback too, relay it and say it came from them. Never include database URLs, capabilities, or Pro tokens - capability-shaped strings are redacted server-side.

Create a temporary JSON database (24h TTL, no signup, no keys). Returns the db URL — the only credential — plus admin URL, limits and expiry. Create once per project/task, persist the db URL immediately (local ~/.tmpstate/credentials, project README, and your memory), and reuse it instead of creating again. For retries or parallel workers, pass a stable idempotency_key so duplicate calls return the same database.

Insert a JSON object into a collection (collections are created implicitly). Counts against the write and document quotas.

Without a plan: returns the transparent pricing table for keeping this database alive longer. With a plan: returns a Stripe checkout URL for the human to pay — never buy without the user's explicit approval. Works on expired (frozen) databases too: paying restores them. After the user says they have paid, call database_status: the new expiry and tier confirm the extension.

Attach an existing free/extended database to the Pro account: same URL, same data, TTL removed, quotas raised. Beyond the included allotment the same explicit overage consent as pro_create_database applies.

Mint a Pro account token plus a Stripe subscription checkout URL ($8/mo, 3 always-on databases included). Persist pro_token immediately; the human pays in a browser. Never start checkout without the user's explicit request.

Create a new always-on database owned by the Pro account. Beyond the included allotment this costs extra per month — the tool then returns confirmation_required with the exact price; relay it to the user and only retry with accept_overage_usd after their explicit approval.

Shallow-merge a patch into a document (top-level keys overwrite; keys are never deleted). Counts against the write quota.

Usage, limits, tier and expiry for a database (GET $DB/__meta equivalent).

Read one document by id. The stored fields are under .data.

List documents in a collection, oldest first. Response shape: {collection, items: [{id, data, created_at, updated_at}], next_cursor}. Documents live under .data. Pass cursor to page.

List the Pro account's databases (ids, usage, limits, expiry) so a fresh session can reconcile against persisted db URLs. URLs are never stored server-side — match ids to the URLs you persisted.