Understand MCP usage and limits
Through Claude, only visibility checks cost credits: 2 credits per platform per prompt, charged after each successful check. Each API key also has request rate limits.
Using Erlin through Claude draws on the same credit balance as the Erlin app, and only visibility checks cost credits. A visibility check costs 2 credits per platform per prompt, and Erlin charges after the check finishes, only for runs that succeed: a platform that fails to return an answer is not charged.
Your Claude usage is separate from Erlin credits: Erlin charges only for the Erlin work Claude asks for.
Review the account before starting work
Checking 10 prompts on ChatGPT, Gemini and Perplexity is 30 runs, so it costs 10 × 3 × 2, or 60 credits, if every run succeeds. This includes the checks /erlin_fanout_audit runs: it adds 8 to 12 prompts and checks each on all three platforms, so a 10-prompt audit costs up to 60 credits.
These cost no credits:
- Reading your dashboard, results, snapshots, competitors and analytics
- Creating, editing, pausing and deleting prompts and collections
- Favoriting competitors
- Syncing Google Search Console or Google Analytics 4 data
- Adding a website, which counts against your plan's website limit instead
If your plan includes bring your own key, checks on the platforms your key covers run on your key instead of your credits. See API keys. Check your balance in Credits.
Handle background work correctly
Visibility checks and data syncs run in the background. Claude starts the work, then checks progress with erlin_get_check_status for checks or erlin_get_sync_status for syncs. Ask Claude to check the status rather than start the same check again: a repeated start creates more runs and costs more credits.
Respond to a limit or error
When credits run out, visibility checks stop until you have credits again. Claude tells you how many credits you have left, how many the check needs, and links you to upgrade. Everything else keeps working.
Erlin also limits how many requests each API key can make:
| Window | Limit |
|---|---|
| Per minute | 60 requests |
| Per hour | 1,000 requests |
| Per day | 10,000 requests |
When you hit a limit, Erlin tells Claude how many seconds to wait. Ask again after that. Normal conversation rarely gets near these limits: large batches of checks are the most likely way to reach them.
Keep a useful record
When you contact Support, include the error Claude reported and the time. Never include your API key. See visibility tools and analytics tools for the status tools.