Search and exports

Federated search

POST /search searches a bounded set of contacts, projects, and tasks. The request requires a term of at least two characters, one to three unique resource types, and an optional limit of at most 50.

search:read is only the federation scope. Every requested type also requires its domain read scope and the normal product permission. Asking for projects and tasks, for example, requires both projects:read and tasks:read. Results never cross the credential’s workspace or cell boundary.

Search is the only newly added AI-facing operation: teamgrid_search is a curated, sensitive MCP tool. Use a dedicated credential and request only the resource types needed for the question.

Bounded asynchronous exports

POST /exports creates a CSV export job for contacts, projects, taskRecurrences, tasks, timeEntries, or auditEvents. Exports are bounded to 10,000 rows and at most 16 selected fields. The request supports archived and updated-at filters and must use a stable Idempotency-Key when retried.

Creating a job requires exports:write; status and download operations require exports:read. The matching domain read scope and analytics permission are also enforced. Poll GET /exports/{id} until the state is succeeded or failed; a successful job records whether the requested result was truncated.

A taskRecurrences export requires task-recurrences:read, tasks:read, and the normal task permission. Its allowlisted projection includes series identity, name, lifecycle status, owner, project, current definition identity/version/hash and timestamps. Optional fields can include the canonical policy, task template and human summary. It exports series state, not occurrence-ledger rows or generated task content; export generated tasks separately with resourceType: "tasks".

Audit exports additionally require audit:read and an immutable createdAtTo snapshot boundary. They may use createdAtFrom, span at most 366 days, and never accept updated-at or archive filters. Only the allowlisted sanitized audit projection can be selected; credentials, raw secrets, request bodies, and internal tenant fields are not exportable.

Exports stay in the workspace’s owning cell and use a dedicated private bucket that is separate from normal file uploads. A completed job record is retained for approximately one hour. Objects under the private developer-exports/ prefix expire after one day, so download a completed export promptly. If the job is no longer available, create a new export instead of retaining or replaying an old download capability.

Header-only download capability

Completed exports use a two-step download flow:

1. Call POST /exports/{id}/download-intent to create a short-lived opaque capability.
2. Call GET /exports/{id}/download with that value only in the X-TeamGrid-Export-Download-Intent header.

curl --fail-with-body \
  --header "Authorization: Bearer $TEAMGRID_API_TOKEN" \
  --header "X-TeamGrid-Export-Download-Intent: $DOWNLOAD_INTENT" \
  --output teamgrid-export.csv \
  "https://api.de.teamgrid.app/v1/exports/$EXPORT_ID/download"

Never put the intent in a query parameter, URL, command history, log, or analytics field. The download endpoint does not accept a query-string fallback. TeamGrid resolves private object storage internally and streams the file through the API with redirects disabled, a 50 MiB response limit, Cache-Control: no-store, and content-type hardening. It does not reveal a storage URL.

Export jobs, intents, metadata, and bulk content are forbidden in every MCP profile.