Skip to main content
The Nanonets Agents Platform Public API lets you trigger an agent run, stream the result, and hand back follow-up answers when the agent needs more from you. The whole surface is six endpoints under /api/v1.

​
What you can do

Run an agent

Start a task with a text query and optional file attachments. Returns a task_id immediately.

Get the result

Poll /tasks/{id}/summary for the final answer, or /result for the full reasoning trace.

Answer the agent

When a task is waiting_for_input, send a follow-up message and let the agent continue.

Update an agent

Hot-patch an agent’s name, system prompt, or full configuration without a redeploy.

​
Where to start

1

Mint an API key

Open the dashboard and create a workspace API key under Settings → API Keys.
2

Run your first agent

Walk through the Quickstart to send your first request.
3

Browse the reference

Every endpoint has a live API Reference page with schemas and a “Try it” panel.

​
Working with the API

Once you’re past the first request, these guides cover the patterns you’ll hit in production:

Async tasks and polling

Polling strategy, terminal vs non-terminal statuses, and a real backoff loop.

Files and structured output

Multipart uploads and shaping the response with output_config.

Interactive tasks

Handle waiting_for_input and reply to mid-task questions.

Errors and pagination

Error envelope, when to retry, and how to walk a cursor-paginated list.

​
Conventions

  • Base URL: https://agents.nanonets.com
  • Authentication: workspace API key, sent as Authorization: Bearer <key>
  • IDs: every entity ID is a UUID
  • Errors: { "error": "<human-readable message>" } with an HTTP status indicating the class
  • Versioning: this is v1. Breaking changes will ship under /v2; /v1 stays stable.