Find who pays for what you sell.
Enter your website. We read what companies publish about the work they pay to get done, and hand back the ones that match, with the words they used. Every row carries the task, the market it belongs to, the companies that published it, the tool they already pay for, how they do it today, a verbatim quote and the link to the page it was published on.
Your contact data tells you who. We tell you why.
Keep the contacts, the CRM and the intent scores you already use. We add the column none of them has: the work that company is paying to fix, in its own words.
- Contact data gives you the name and the email. We give you the task they published, with the date.
- Your CRM tells you what you sold them last. We tell you what they are hiring to fix next.
- An intent score tells you something moved. We show you the sentence that moved it.
An intent score says they looked. A posting says they hired.
When a company decides a job has to get done, it writes the job down and publishes it. That is an approved budget in plain language, not a probability someone modelled. We read those postings the day they appear, turn them into rows, and keep the sentence and the link on every one.
One record of the work companies pay to get done
- Look up a company. Everything one company is paying to get done, with the tool they already use and how they do it today.
- Find who pays for a job. Every company paying to get the same work done, with the date each one published it.
- Watch the work move. What appeared, what spread, and what is being industrialised. The difference, not the pile.
- Check any of it. Every field comes with the words it came from and a link to the page it was published on.
- Feed your pipeline. MCP for the agents, REST for the jobs, CSV for the spreadsheet. The provenance travels with it.
Everything you need to put it in production
- Every row is a company and a sentence. The task they need done, the companies that need it, the tool they already pay for, and how they do it today, each field with the words it came from and a permalink.
- Ask in your own words, get the receipts. No query language and no filter builder. Ask the way you would ask an analyst, and the answer arrives with the quotes and the links attached.
- It lands in your pipeline, not on another dashboard. Fourteen tools over MCP for the agents, a REST API for the jobs that run unattended, CSV for the spreadsheet that runs the quarter. Add it with
claude mcp add --transport http echosift https://mcp.echosift.io/mcp. - Nobody has to take your word for it. Every quote is checked against the original text by machine before it becomes a row, and a row that is not fully backed says which part is thin and why, in words.
- Live, on a clock you can check. The whole set is rebuilt every night and in the morning you get the difference, not the pile again. Markets carry their own coverage, the empty ones marked.
Start free. Pay when it runs your pipeline.
Free: ask in your own words and get answers with the verbatim quotes and the links, briefs and product comparisons, the MCP server for Claude, Cursor and Codex, the market map and coverage including the empty markets, the morning difference, and your own context so answers come back in your terms. No credit card and no trial clock, just a one-time allowance you can see inside the product.
Pro, $199 per month or ten months when you pay for the year: the whole corpus with no ceiling, every market followed and rebuilt nightly, the REST API with scoped keys, webhooks, the weekly digest and Telegram alerts, export of any market or brief as CSV, JSON or Markdown with the provenance attached, organisation search by name, the why-now read on any task, and market analytics. Complete means complete: no blurred fields, no export lock, and no per-record billing. You are never charged for looking. See pricing.
Questions, answered
- What exactly is a row?
- A recurring operational task, with the market it belongs to, the companies that published it, the tool they already use, how they do it today, a verbatim quote and the link to the page it came from. Seven columns, filled or honestly empty.
- How is this different from intent data?
- Intent data gives you a score you cannot check. We give you the sentence and the link, so your own reviewers can verify a row before your team acts on it.
- Where does the data come from?
- Public job boards of companies that buy tools, and public discussions where operations people talk: developer issues, Stack Exchange, vendor communities, Hacker News. Nothing is licensed from a reseller and nothing is scraped from behind a login.
- Can my agent use it directly?
- Yes. The MCP server exposes the same rows and the same evidence to Claude, Cursor, Codex or any MCP client, with OAuth or a scoped API key.
- What happens when a market is thin?
- It says so. Coverage is published per market including the zeros, and a task that has not crossed three named companies stays out of the list rather than padding it.
- What do I get for free?
- The real product, not a demo of it: ask in natural language, read the evidence with the links, and use the MCP server. There is a one-time allowance rather than a trial clock, and you will see exactly where you stand against it inside the product.
- Who is this not for?
- If you need contact details, emails or phone numbers, this is not it and we will not pretend otherwise. We carry the work a company published and the words it used, not the people. Bring your own contact data and we add the reason to reach out.
- How fresh is a row?
- A public posting reaches the corpus within 48 hours of being published, and the whole set is rebuilt every night. Every row carries the date each company published the work, so you can judge freshness yourself instead of trusting a label.
- Why a flat month and not per record?
- Per-record billing punishes you for looking. Providers in this category charge per call or per credit, which means the cost of checking a market is unpredictable before you check it. One monthly price, every row, no meter on curiosity.
- How do you handle the people in the data?
- We track the work, not the person. Author handles are pseudonymised at ingestion and the original is never written to disk. We quote the claim, never the human.
Ask it something you already know the answer to.
Natural language in, cited evidence out. Free to start, no card, and the links are there from the first answer. Start free or sign in or read the EchoSift blog, including what an MCP server is, how to find startup ideas and how to do customer discovery. Built in public by @alepbuilds.