REST API

Assistants

Where to get assistant_id and llm_model_id: the list of assistants the token owner can see, and the models of each one.

Listing assistants

GET /api/assistants

Returns the same assistants the token owner sees in the Aist Ai catalog: the user's own assistants plus the catalog of their organization, bundle or plan. id is the assistant_id you pass when creating a task.

curl https://agent.aist-ai.com/api/assistants \
  -H "Authorization: Bearer $AIST_AGENT_TOKEN" \
  -H "Accept: application/json"
{
  "data": [
    {
      "id": 3,
      "slug": "writer",
      "title": "Editor",
      "description": "Writes and edits texts",
      "avatar_url": "https://…",
      "owned": false,
      "categories": ["Work", "Creativity"]
    }
  ]
}
FieldDescription
idassistant_id for POST /api/tasks.
slugKey for requesting the assistant's models (see below).
title, description, avatar_urlDisplay data — same as in the catalog.
ownedtrue — the assistant was created by the user.
categoriesCatalog categories. Each assistant appears once, even if it belongs to several.
Hidden assistants are not listed
The list matches what the user sees on the website. Assistants that aren't in the catalog (internal ones, built for integrations) don't show up here — the platform administrator gives you their assistant_id along with access. You create tasks with such an ID as usual.
Code execution is not available through Agent yet
Assistants called through Agent don't run code: even if an assistant can do it in the catalog, it won't have that tool in a task, and code_execution in options is false. Reading spreadsheets and documents works as usual.
Video works like images
If an assistant has a video generator, the clip is generated right inside the task and arrives as a file item in output. Unlike the chat on the website, Agent has no confirmation step.

Assistant models

GET /api/assistants/{slug}

Pass models[].id to a task as llm_model_id. Without it, the assistant answers with the model marked is_default: true. An assistant the token can't access returns 404.

{
  "id": 3,
  "slug": "writer",
  "title": "Editor",
  "description": "Writes and edits texts",
  "avatar_url": "https://…",
  "models": [
    { "id": 11, "label": "Fast", "description": null, "is_default": true, "features": [] },
    { "id": 12, "label": "Deep", "description": "Thinks longer", "is_default": false, "features": [] }
  ],
  "options": { "internet_search": true, "read_documents": true, … }
}
# A task for a specific model of the assistant
{ "assistant_id": 3, "llm_model_id": 12, "input": [{ "type": "text", "text": "…" }] }

Errors

CodeWhen
401The token was not accepted, has expired or was revoked.
404No assistant with this slug, or the token can't access it.
502Aist Backend is temporarily unavailable — try again later.

The same without writing code — in the dashboard: paste your token, and the assistants with their models are on the Assistants tab.