Prospect Data You Can Wire Into Your Own Pipeline

Yes. Cleanlist REST API v2 sits at https://api.cleanlist.ai/api/v2 with around 30 endpoints grouped into workspace and identity, search, lead lists, enrichment, smart agents, and sync and export. Authentication is a clapi_ Bearer token minted from the dashboard with no approval queue and no sales call, and every endpoint checks one of 14 granular OAuth scopes, so a key issued for reads cannot start a paid enrichment. There is a published OpenAPI 3.x spec. Programmatic access requires Pro at $229/month or above, or an AppSumo Tier 5+ license. Call GET /whoami first to confirm the identity, scopes, plan tier and organization a key resolves to before wiring it into a job.

Cleanlist allows 60 requests per minute per organization and 30 requests per minute per API key. People Search carries a separate hard cap of 60 searches per UTC day per API key, and that is the limit most agent workloads hit first, so plan pagination and caching around it rather than around the per-minute number. Exceeding any limit returns HTTP 429 with a Retry-After header and a rate_limited error body, which is enough for a worker to back off deterministically. AppSumo lifetime licenses without a paid plan are held to 30 requests per minute per organization. Cleanlist publishes these numbers rather than leaving you to discover them in production.

No. The current v2 Cleanlist API is poll-based. Asynchronous enrichment returns a workflow_id and you poll it until the status is terminal, with exponential backoff as the recommended pattern. Synchronous operations (company enrichment, search, CRM sync, exports) answer in the same request and need no polling at all. There are no v2 webhooks, so a scheduled job that enriches a list needs a second scheduled step that polls for completion rather than a listener waiting on a callback. Older Cleanlist tutorials that describe receiving a webhook are documenting the frozen v1 surface, not v2.

Yes, through a generic HTTP node. There is no native Cleanlist app in the n8n, Make or Zapier directories. Point an HTTP step at https://api.cleanlist.ai/api/v2, send your clapi_ Bearer token, and handle completion in a second scheduled step, because the API has no v2 webhooks to call your workflow back. For HubSpot, Salesforce and Outreach you can often skip the automation platform entirely: POST /sync/crm pushes leads directly at 0.2 credits per lead, and POST /sync/sequencer pushes to Lemlist. Cleanlist does not send email, run sequences or dial numbers itself, it hands the record to the tool that does.

Cleanlist gates paid bulk work behind a signed quote. You call POST /credits/estimate, get back an HMAC-signed quote_id with a 5 minute TTL, and pass that quote_id on the request that actually spends, so a run cannot exceed the quote it was signed against and a stalled agent simply lets the quote expire. Search costing 0 credits means the exploratory half of an agent session is free by construction, and pay-for-results billing settles reserved credits back on records the waterfall cannot resolve. The Cleanlist MCP server enforces the same quote mechanism against the same wallet, so the cap holds whether the caller is your code or Claude.

No. The Cleanlist REST API takes structured filters (job title, seniority, department, headcount, industry, location and similar), not a free-text prompt, and there is no natural-language search endpoint on it. Natural language lives one layer up: the Copilot inside the Cleanlist app, and the MCP server, where the model turns your instruction into the same structured filter payload a script would send. If you want an agent that accepts plain-English list requests, connect the MCP server and let it compose the filters. If you are writing a script, build the filter object yourself against the published OpenAPI spec.

A verified email is 1 credit, so 10,000 leads is 10,000 credits, which sits inside Scale at $599/month for 15,000 credits. Adding a direct dial takes each record to 11 credits, or 110,000 credits for the same file, which is past any published plan allowance, so scope the phone pass to the accounts you will actually dial rather than the whole list. Cleanlist bills for results, so records the waterfall cannot resolve settle their reserved credits back and real spend usually lands below the worst case. Quote the job with POST /credits/estimate before dispatching it and the returned quote_id caps what that run can spend.

Cleanlist people and company search runs on firmographic and role attributes: job title, seniority, department, headcount, industry, location and similar, 20+ filters in total. There are no intent-data, technographic, revenue-range or funding-stage filters, and no job-change detection or alerting. If your scoring model depends on one of those signals, source it separately and join on the company domain, or narrow the segment with the filters Cleanlist does expose and let a smart agent (0.5 to 3 credits per lead) research the rest on only the records that survive the filter pass. Naming the gap up front is cheaper than discovering it after you have written the integration.