Credits and rate limits

The two meters

Meter	What it counts	Spent by
API credits	Calls. Most cost 1, heavier ones cost 2.	Every REST request and every MCP tool call.
Export rows	Rows. An export costs exactly the rows it contains.	Creating an export, and the Export buttons on dashboard cards. Nothing else.

Both belong to the project owner. Both are shown in Settings, API usage. GET /me returns both balances and is free. The Export buttons on dashboard cards spend Export rows too, one per row in the file. Before anything is charged you see the row count and confirm. The file downloads at once, and a copy is kept for days under Project Settings, Export data, where everyone on the project can download it. A card file stops at a number of rows set by your plan (the “Rows per card export” column below); only the rows written are charged. For more rows, use Export data.

What each call costs

Row endpoints return up to 25 rows per call and a next_cursor; each call is one credit, so 350 responses is 14 calls. Running a prompt now also spends 1 realtime credit. Writing an article also spends 1 AI content credit. The API credit is only for the call itself.

Every response tells you

"credits": { "charged": 1, "remaining": 2948 }

The same numbers are in the headers X-RadarKit-Credits-Charged and X-RadarKit-Credits-Remaining.

Failed calls cost nothing

A call is charged before it runs and refunded if it fails on our side. Validation and permission errors are refused before anything is charged. With no API credits left, every metered call gets 402 INSUFFICIENT_CREDITS. With no Export rows left, creating an export gets 402 INSUFFICIENT_EXPORT_ROWS.

Plan allowances and speed

The plan bucket refills on every renewal. Top-up credits do not expire and are used after the plan bucket is empty. Both rate limit windows must have room. Enterprise plans have custom allowances and rate limits. Contact us for yours. Need more? Write to mail@radarkit.ai.

Extra limits

Calls	Limit per key
Heavy reads: top source domains, share of voice, sentiment, export estimates	per minute
Creating an export, running a prompt now, creating an article	per minute
Adding topics or prompts, retrying an article	per minute
Prompts added	per rolling 24 hours, REST and MCP together
Any single IP address	per minute, whatever keys it uses

When you hit a limit

You get 429 RATE_LIMITED and a Retry-After header with the seconds to wait. Nothing was charged.

{
  "error": {
    "code": "RATE_LIMITED",
    "message": "Too many requests. Slow down and retry.",
    "details": { "retry_after": 21 }
  },
  "request_id": "req_01j9x4r2s3t4u5v6w7x8y9z0a1"
}

Every response carries X-RateLimit-Limit and X-RateLimit-Remaining for the per-second window.

Ask for the largest page you need, up to rows. Use the summary endpoints before the row endpoints. For more than a few thousand rows, create an export: one call, one file.

Overview

Account

Projects

Topics

Prompts

Visibility

Responses

Sources

Competitors

Rankings

Query fanout

Web search

Sentiment

Exports

Content

Changelog

Credits and rate limits

The two meters

What each call costs

Every response tells you

Failed calls cost nothing

Plan allowances and speed

Extra limits

When you hit a limit

​The two meters

​What each call costs

​Every response tells you

​Failed calls cost nothing

​Plan allowances and speed

​Extra limits

​When you hit a limit

The two meters

What each call costs

Every response tells you

Failed calls cost nothing

Plan allowances and speed

Extra limits

When you hit a limit