2,500 free credits — no credit card required. Start building
Every business in a niche and city, with phone, email, and website already resolved. You pay for the rows you get back — no per-call minimum, no subscription, credits that never expire.
16 endpoints · 3 sources · 2,500 free credits
SenseCollectcurl "https://sensecollect.com/v1/maps/search?query=dentists&location=Berlin&limit=50" \ -H "x-api-key: $SENSECOLLECT_KEY"{ "success": true, "status": "completed", "source": "google_maps", "credits_used": 400, "credits_remaining": 9600, "count": 50, "data": [ { "id": "ChIJ0xT4…", "type": "place", "title": "Mitte Dental Studio", "url": "https://maps.google.com/…", "contact": { "phone": "+49 30 000000", "email": "hello@mittedental.example", "website": "" }, "location": { "address": "Torstraße 1, Berlin" }, "metrics": { "rating": 4.6, "reviews": 42 } } ]}Collecting from the public surfaces your product needs
The usual complaint about scraping platforms is the invoice, not the data. This one charges per row, tells you the cost in the response, and refunds itself when a run fails.
No per-call minimum, no compute units. Credits are held, then settled against the rows that came back — ask for 200, get 12, pay for 12. Failed runs refund themselves and every response reports what it spent.
See what a row costsEvery record arrives in the same top-level shape — author, contact, location, metrics, content. No per-source glue code, no schema that drifts when a page does, and the raw payload is still there under `raw`.
See the schemaAn MCP server at /v1/mcp puts every source in front of an agent on the same key. Or use plain GETs with a machine-readable catalog at /v1/sources — either way the agent can budget itself before it spends.
See agent setupAgencies and freelancers use it to build lead lists their clients pay for. Reviews and trend data sit on the same key when the research work comes up.
One call returns businesses for a niche and city with phone, email, website, rating, and review count already resolved — leads, not listings. Filter to the ones missing a website and you have a web-design lead product.
GET /v1/maps/no-website?query=restaurants&location=Hamburg
Pull reviews for any public ASIN and pipe the text into sentiment scoring, objection mining, or a support knowledge base. The complaints-only endpoint skips the paging.
GET /v1/amazon/complaints?asin=B09G9FPHY6
67% of records are complaints
Poll a hashtag on a schedule and diff the engagement counts. Plays, likes, comments, and shares arrive as numbers, so velocity is a subtraction rather than a parse.
GET /v1/tiktok/hashtag?hashtag=fitnesscoach
Run the same question against maps, reviews, and posts, then join on the unified record shape. One schema means one join key and one storage table instead of three.
GET /v1/maps/search?query=coworking&location=Lisbon
Click any tile to see its parameters, copy the request line, and run it live against the API right here.
/v1/maps/search· MapsBusinesses matching a query and location, with phone, email, website, rating, and review count.
curl "https://sensecollect.com/v1/maps/search?query=dentists&location=Berlin&limit=50&verify=true" \ -H "x-api-key: $SENSECOLLECT_KEY"// Hit “Run it” for a live response. Data endpoints// need your own key, so this one answers 401 until// you send x-api-key.querystringrequiredlocationstringlimitintegerenrichbooleanverifybooleanpersonal_onlybooleanskip_claimedbooleanA place, a review, and a post differ in what they mean — not in how you read them. Write the parser once.
id, type, source, title, url, content, published_at, author, contact, location, metrics. Present on every record from every endpoint, empty rather than missing when a source has nothing to say.
Everything under `metrics` is a number or null — ratings, review counts, plays, likes. No string parsing before you can average, sort, or diff.
`raw` carries the full source-shaped record. Use the unified fields for the 90% case and drop into `raw` when you need something source-specific.
{ "id": "ChIJ0xT4…", "type": "place", "source": "google_maps", "title": "Mitte Dental Studio", "url": "https://maps.google.com/…", "content": "Dentist", "published_at": null, "author": { "name": "", "handle": "", "url": "" }, "contact": { "phone": "+49 30 …", "email": "hello@…", "website": "" }, "location": { "address": "Torstraße 1, Berlin" }, "metrics": { "rating": 4.6, "reviews": 42 }, "raw": { … }}{ "id": "https://tiktok.com/@coach.mara/video/…", "type": "post", "source": "tiktok", "title": "3 things I stopped doing…", "content": "3 things I stopped doing…", "published_at": "2026-05-02T09:14:00.000Z", "author": { "name": "coach.mara", "handle": "coach.mara", "url": "…" }, "metrics": { "plays": 812400, "likes": 94100, "comments": 2210 }, "raw": { … }}Every source you add on your own is another proxy pool, another retry policy, and another schema to keep in sync. Behind one key, that is somebody else's on-call rotation.
3 vendors · 3 schemas · 3 bills
GET /v1/maps/search
GET /v1/amazon/reviews
GET /v1/tiktok/hashtag
→ same envelope
→ same record shape
→ same credit ledger1 key · 1 schema · 1 bill
Plug SenseCollect into Claude, Cursor, Codex, Windsurf — any MCP-compatible client — in seconds. Or point your agent at the quickstart and it onboards itself.
One JSON block and every endpoint becomes a first-class tool. Same key, same per-row pricing, same records — the MCP server dispatches to the handlers the HTTP API uses.
{ "mcpServers": { "sensecollect": { "type": "http", "url": "https://sensecollect.com/v1/mcp", "headers": { "x-api-key": "sc_live_YOUR_KEY" } } }}The integration guide is markdown at a stable URL — auth, every endpoint, the response schema, the error codes, and the rules worth keeping in project context. Your agent fetches it and wires itself up.
curl -s https://sensecollect.com/agent-quickstart.mdEvery source, endpoint, parameter and per-row price as JSON. An agent plans and prices a call before anyone pays for anything — then reads credits_used off each response to budget itself.
curl -s https://sensecollect.com/v1/sourcesMaps, Amazon, TikTok
One schema · one auth header
HTTP and MCP, same key
No card. Around 147 enriched leads.
No subscription, no monthly minimum, and no charge for a call that comes back empty. Buy another pack only when you need one.
You pay for rows, not requests — so tell us how many rows a month you need. Calls that return nothing cost nothing.
Agency Pack
80,000 credits · $99 · never expire
Estimates use published per-row rates. Every response reports the credits it actually consumed, and failed runs are refunded in full.
Wiring up the API and validating a niche
10,000 credits
$1.90 per 1,000 credits · 588 enriched local leads
Weekly prospecting runs and agent workloads
30,000 credits
$1.63 per 1,000 credits · 1,764 enriched local leads · 14% cheaper per row
High-volume list building across a client roster
80,000 credits
$1.24 per 1,000 credits · 4,705 enriched local leads · 35% cheaper per row
Everything about credits, refunds, and what a call actually costs.
No. SenseCollect sells one-time credit packs. You buy a pack, spend credits when you call the API, and top up when you choose to. There is no monthly fee, no seat count, and nothing to cancel.
No. Credits stay on the account until you spend them, whether that takes a week or a year.
Calls are priced by the rows they return — there is no per-call minimum. An enriched local lead is 17 credits, an Amazon review 7, a TikTok post 10, and a credit is a tenth of a cent at list price. The catalog at /v1/sources publishes the per-row rate for every endpoint, and every response reports exactly what it consumed.
Yes — add verify=true to a Maps search. Every address is checked before the row is returned, rows without a deliverable contact are dropped, and only the rows that pass are charged. Add personal_only=true and shared inboxes like info@ are dropped too, so you get a named person or nothing.
You are charged for what came back. Credits are held against your limit when the run starts, then settled against the rows it collected and the difference is refunded — so asking for 200 places in a town that only has 12 costs you 12 rows, not 200. A call that finds nothing is free.
No. If an upstream run fails, the credits are returned to your account automatically and the response tells you they were refunded.
New accounts start with 2,500 credits, no card required. /v1/account, /v1/sources, /v1/jobs/{id}, watches, and claimed leads are always free to call.
3 sources. One API key. One schema. Two lines to your first response.
curl "https://sensecollect.com/v1/maps/search?query=dentists&location=Berlin" \ -H "x-api-key: $SENSECOLLECT_KEY"Most disappointment with business data tools comes from expecting fields that were never on the source. Here is the honest inventory.
Read articleGoogle Maps · 9 min readThree surfaces look identical on a results page and expose completely different fields. Knowing which one you are looking at decides whether your data is collectable.
Read articleLead generation · 6 min readA personal injury firm and a yoga studio are the same API call with a different query string, and completely different businesses to sell to.
Read articleStill unsure? The catalog at /v1/sources is public — read it before you sign up.
An HTTP API for local business lead data, with Amazon reviews and TikTok posts on the same key. You send a GET request with an API key; SenseCollect runs the collection against the live source and returns records in a single unified schema, priced by the rows it returns.
Yes — there is an MCP server at /v1/mcp exposing every endpoint as a tool, on the same key and the same per-row pricing. Agents that do not speak MCP can use the plain GET endpoints and read /v1/sources to discover them.
For a single source, building it yourself is reasonable. What you own afterwards is the source changes, the retries, the proxying, the deduping, and a schema that drifts every time a page does — SenseCollect is the version where that is someone else's on-call rotation.
Every call collects against the live source at request time. Nothing is served from a stale cached index, which is also why a large pull takes seconds rather than milliseconds.
Small pulls typically return inside one request. Larger ones can outlive the request budget — those return a job id with a 202, and you poll /v1/jobs/{id} for the result. Pass async=true to skip waiting entirely.
There is no per-minute request cap. Concurrency is bounded per account instead, and credits are the real limit — if you need dedicated throughput, get in touch before you build against it.
Yes. Every API run is also saved in your dashboard as a downloadable CSV, so the same call serves a pipeline and a spreadsheet.