Quickstart HTTP

Before wiring manual HTTP calls, use AI Platform -> Setup to copy the correct organizationId, create or fetch a token, and run the public setup verification check.

Call the API over HTTPS, or http://localhost in dev. Base URL is your deployed saas-ai-api host.

Prerequisites

  • organization ID for your tenant
  • user access token for access_token_only routes
  • optional API key for public routes that allow api_key_or_access_token

List skills

GET /api/skills?organizationId=<orgId>
Authorization: Bearer <accessToken or apiKey>

API keys on this route need skills:read.

Run a skill

POST /api/skills/<skillId>/run
Authorization: Bearer <accessToken>
Content-Type: application/json

{
  "organizationId": "<orgId>",
  "skillVersionId": "<optional version id>",
  "variables": {}
}

This route is access_token_only. The response matches skillRunResponseSchema and includes runId, output, validation, and diagnostic.

Run an agent and fetch stable public run detail

POST /api/agents/<agentId>/run
Authorization: Bearer <accessToken>
Content-Type: application/json

{
  "organizationId": "<orgId>",
  "versionId": "<optional version id>",
  "variables": {}
}

Response matches agentRunResponseSchema.

Then fetch persisted run detail:

GET /api/agent-runs/<runId>?organizationId=<orgId>
Authorization: Bearer <accessToken or apiKey>

API keys on this route need runs:read.

See also

Was this page helpful?