Skip to content
Hyperresearch docs

Vault

Vaults, reports and citations.

Browse recent notes or search with FTS, semantic or hybrid mode. A note keeps its source, producing run, lifecycle state and rights metadata. Report citations resolve to source notes for provenance.

Terminal window
curl "https://api.hyperresearch.ai/v1/vault/search?q=grid+storage&mode=hybrid" \
  -H "Authorization: Bearer <YOUR_API_KEY>"

Vault export is an asynchronous tar.gz job. Poll GET /v1/jobs/{job_id}, then use its short-lived download_url. The console’s direct API import sends tar.gz bytes with Content-Type: application/gzip.

Large agent transfers

Section titled “Large agent transfers”

The CLI streams exports into a private temporary file before safe extraction. Agent imports use resumable 8 MiB parts: create an upload at /mcp/uploads/vault-imports, PUT ordered parts with X-Content-SHA256, then complete or delete the upload. The CLI handles these routes and resumes an open upload job; clients should not store its bearer token.

PayGo accepts up to 5 GiB and Enterprise up to 50 GiB compressed and expanded, with 25,000 entries, 5 GiB per archive entry, 25 MiB per note, three pending imports per workspace, and a 24-hour upload expiry.

Further reading

Section titled “Further reading”
  • The vault: how sources accumulate across runs and what a run reads first.
  • Trust: retention, isolation between workspaces and subprocessors.