DoubleX AgentesDocs

API reference › Resources

Creations

List and download the files agents created on their computers.

GET/api/artifacts#

Lists the files of the agents you can see, plus the ones shared with you, most recent first.

QueryDescription
agentIdOnly one agent's files.
curl "https://doublex.ai/api/artifacts?agentId=agt_8f3a2b1c9d0e" -H "Authorization: Bearer $DOUBLEX_API_KEY"
JSON
{
  "artifacts": [
    {
      "agentId": "agt_8f3a2b1c9d0e",
      "agentName": "CRM Analyst",
      "path": "relatorios/pedidos-outubro.md",
      "kind": "documento",
      "size": 4821,
      "modifiedAt": "2026-10-27T09:12:00.000Z",
      "share": { "scope": "agent" },
      "canEdit": true
    }
  ]
}

kind can be documento, planilha, apresentacao, dashboard, codigo or outro.

share.scope says who can see the file: agent (internal: whoever sees the agent), team (with teamId), organization or public (with url, the public link path). canEdit tells whether you can edit the agent, and therefore share and delete the file.

GET/api/artifacts/file#

Returns a file's content.

QueryDescription
agentIdRequired.
pathRequired. The path shown in the list.
downloadIf present, the response comes as an attachment (Content-Disposition: attachment).
curl -G https://doublex.ai/api/artifacts/file \
  --data-urlencode "agentId=agt_8f3a2b1c9d0e" \
  --data-urlencode "path=relatorios/pedidos-outubro.md" \
  -H "Authorization: Bearer $DOUBLEX_API_KEY" -o pedidos-outubro.md

The Content-Type follows the file extension (text/html, text/csv, application/json, text/markdown, image/svg+xml, .xlsx spreadsheets, application/pdf; everything else is sent as text). HTML pages are served in an isolated sandbox, with no network access. Text files can be up to 2 MB; .xlsx spreadsheets and PDFs, up to 10 MB.

GET/api/pdf#

Creates an A4 PDF from a .md document, an .html page or a .csv table in the agent's workspace, or from a live creation with its current data. It requires the same permission as viewing the file or the creation; each new PDF uses credits like the cloud browser, and the same content comes from the cache at no cost.

QueryDescription
agentId + pathA file from the list.
liveOr the id of a live creation (live_…).
curl
curl -G https://doublex.ai/api/pdf \
  --data-urlencode "agentId=agt_8f3a2b1c9d0e" \
  --data-urlencode "path=relatorios/pedidos-outubro.md" \
  -H "Authorization: Bearer $DOUBLEX_API_KEY" -o pedidos-outubro.pdf

DELETE/api/artifacts/file#

Deletes a file from the agent's computer. Takes agentId and path in the query and requires edit permission on the agent.

Response: { "ok": true }. Deleting the file also turns off its public link.

PUT/api/artifacts/share#

Changes who can see a file. Requires edit permission on the agent.

FieldDescription
agentIdRequired.
pathRequired. The path shown in the list.
scopeRequired. agent (internal), team, organization or public.
teamIdRequired with team. You must be a member of the team.
curl
curl -X PUT https://doublex.ai/api/artifacts/share \
  -H "Authorization: Bearer $DOUBLEX_API_KEY" -H "Content-Type: application/json" \
  -d '{"agentId":"agt_8f3a2b1c9d0e","path":"dashboards/funil.html","scope":"public"}'
JSON
{ "share": { "scope": "public", "url": "/compartilhado/Q39iOloCAPMHVs4giIb69TxY" } }

Team and organization open only the file, not the agent. With public, anyone with the link sees the file without signing in, always the latest version; the page and the raw content (/api/public/artifacts/<token>, with ?download to download) are not indexed by search engines. Switching from public to another scope turns the link off; publishing again creates a new link.