REST API

When to use the REST API

The MCP server is the recommended way to connect AI agents to Pendium — it handles tool discovery, parameter validation, and structured responses automatically.

But if you're building a custom integration, scripting with curl, or working in an environment that doesn't support MCP, the REST API gives you direct HTTP access to the same data.

Base URLs:

https://pendium.ai/api/visibility   # visibility scans, reports, topics, personas
https://pendium.ai/api/v1/site      # site content, for rendering on your own front end

Authentication

Pass your API key in the x-api-key header:

curl https://pendium.ai/api/visibility/account \
  -H "x-api-key: pendium_xxxxxxxxxxxx"

Or, for HTTP clients that only support bearer auth (Databricks Unity Catalog, Bruno, Insomnia, n8n), use Authorization: Bearer:

curl https://pendium.ai/api/visibility/account \
  -H "Authorization: Bearer pendium_xxxxxxxxxxxx"

No key? Most endpoints return a 401 with instructions on how to sign up. The score lookup endpoint works without authentication.

An account in read-only mode can read its data through the API but can't trigger writes or spend — write/scan endpoints return a 403 (code: "READ_ONLY"); read endpoints (polling, reports, lookups) work normally. A free-plan API key can't trigger scans at all: POST /api/visibility/scan returns a 403 with code: "PLAN_UPGRADE_REQUIRED". A paid key rescanning a group-shared brand whose owner is on the free plan gets a 403 with code: "OWNER_PLAN_UPGRADE_REQUIRED" until the owner upgrades; polling that brand's existing scans still works. Separately, during a platform upgrade we may briefly pause all new scans: POST /api/visibility/scan then returns 503 (code: "SCANS_PAUSED") while polling and reports keep working.

See Authentication for full details.

Endpoints at a glance

Scanning and reports

A scan only creates a public Brand Index page when you trigger it with publishToBrandIndex: true. Once a brand has a page, it shows the newest scan.

Scans report transient failures honestly: if a run can't reach any AI engine (a temporary hiccup), it comes back with scanFailed: true (when polling the latest completed scan) or success: false (in a webhook callback, which is scoped to the run that just finished) and a 0 score — so you never mistake a transient zero for a real result. The same honesty applies per platform: any platformScores entry the scan didn't query comes back with queried: false and a null score, not a 0. Which platforms a scan queries, and its query cap, follow the brand owner's plan rather than the calling key's; see Platforms by plan.

Site content

The site export is a versioned contract for your own renderer, so its responses carry an ETag and no _pendium block. Use your own front end walks through a site built on it.

Brand data

Topics and personas

Response format

Every Visibility API response includes a _pendium object with context about what you're looking at and what to do next. This is designed for both humans reading JSON and AI agents parsing responses.

{
  "plan": "starter",
  "creditBalance": 450,
  "agents": [...],
  "_pendium": {
    "what_is_pendium": "Pendium monitors how AI platforms perceive brands...",
    "what_you_are_seeing": "Your account overview with 3 brand agents...",
    "next_steps": ["Trigger a scan with POST /api/visibility/scan..."]
  }
}

Rate limits

• Authenticated: 60 requests/minute, plus a shared 5/minute for POST to scan, scan-preview, topics, and personas
• Unauthenticated: 15 requests/minute (keyless POSTs to scan, topics, and personas count against the same 15)

Limits are counted per server instance, so they're approximate.

See Rate Limiting for details.