Launching 1 December: Firmable Data API.Data API launches 1 Dec. Get beta access

ProductsFirmable Decisions

Give your agent answers it can act on.

Firmable Decisions answers yes or no questions about companies and people, from Firmable's own data. Every answer comes with a confidence score and the facts behind it, and with reasoning on, the source and date too.

  • Qualify every account against your ICP
  • Clean stale contacts out of your CRM
  • Check an email is likely to deliver before you send
  • 1 cent per answer
  • 34 checks
  • Unknown answers free
Sample response for
Is Shopify publicly listed?
{
  "success": true,
  "tool": "is_publicly_listed",
  "result": {
    "decision": "yes",
    "confidence": 1,
    "details": {
      "list_loaded": "Nov 2026",
      "exchange": "Nasdaq and TSX"
    },
    "entity": {
      "type": "company",
      "id": "f000000138272",
      "name": "Shopify"
    },
    "data_as_of": "2026-11-11",
    "reasoning": {
      "reason": "Shopify is publicly listed: on the Nasdaq and TSX register, as of Nov 2026.",
      "source": "Firmable registers",
      "date": "2026-11-11",
      "would_change_if": []
    }
  }
}

Real responses, trimmed.

Try it

Ask about a company you know

Real answers from Firmable, recorded 17 Nov 2026. Sign up to ask about your own accounts.

Firmable DecisionsStripe
Question
Is it publicly listed?
Decision
no confidence 1.00
Reason
Stripe is on none of the registers checked (ASX, NZX, Hong Kong Stock Exchange, Tokyo Stock Exchange, Singapore Exchange, Bursa Malaysia, Indonesia Stock Exchange, Philippine Stock Exchange, SET, HOSE, KRX, Nasdaq, NYSE, NYSE American, TSX, CSE), as of Nov 2026.
Source
Firmable registers, 2026-11-11

Run it on your own accounts with $25 in free credit.

Sign up free

Decisions in 30 seconds

One question for every record you hold.

A domain, a job title, a work email or a mobile number each raises a question before your agent acts: is the account a fit, would this person buy, are they still in the role, will the email land, can we call? Decisions answers each one with yes, no or unknown, and with reasoning on, the reason why.

  • is_icp_fit
  • is_decision_maker
  • is_contact_current
  • is_email_valid
  • is_mobile_on_dnc
Animation: CRM records become questions. Is this company an ICP fit? Would this person make the buying call? Is this CRM contact still current? Will this work email deliver? Is this mobile on Do Not Call? Each is answered with the tool that checks it.

15 times fewer tokens for the same ten answers

We asked a web-search agent and a Firmable agent whether ten well-known companies are publicly listed. Press run, then click any answer to see the fact behind it.

Measured on 14 November 2026

Are these ten companies publicly listed?

    Web: plain text of each Wikipedia article, about four characters a token. Firmable: all ten responses.

    Run this on your own accounts. Your free credit covers the first test.

    The same question, with and without Decisions

    The web finds a Wikipedia page and hedges. Firmable checks its exchange lists and answers, with the date.

    Your agent
    1. Is Grab publicly listed? This campaign is for listed companies only.
      • Searched the web for “Grab stock exchange listing”
      • Read en.wikipedia.org/wiki/Grab_Holdings
    2. According to Wikipedia, Grab trades on the Nasdaq. I can't tell how up to date that page is, so you may want to check before relying on it.

      About 3,700 tokens read. No date on the answer.
    3. Use Firmable Decisions instead.
      • make_decision is_publicly_listed grab.com
      decision: "yes"   confidence: 1
      reason: "Grab is publicly listed: on the Nasdaq register, as of Nov 2026."
      source: "Firmable registers"   date: "2026-11-11"
    4. Yes. Grab is on the Nasdaq register, checked on 11 November 2026. I've kept it in the campaign.

      About 240 tokens. Source and date included.

    The open-web reply is an illustration. The token counts and the Firmable response are real, measured on 14 November 2026.

    How it works

    Ask, then check the record, then act.
    1. 1

      Ask

      Name a tool and who it is about. Through MCP, you can ask in plain English instead.

    2. 2

      Check the record

      Firmable runs the right tool against its record. No web search.

    3. 3

      Act

      Branch on yes, no or unknown. Act above your confidence floor.

    decision
    Yes, no or unknown.
    confidence
    How sure the answer is, 0 to 1.
    details
    What was checked and found.
    entity
    Who it's about, and how matched.
    data_as_of, version
    Data date and tool version.
    reasoning
    With reasoning on: reason, source, date, would_change_if.

    Let confidence decide what happens next

    These are eight real ICP checks. Move the threshold to see which accounts an agent would act on, skip, send to a person or look up first.

    Real is_icp_fit answers, 14 November 2026

    Move the threshold to see where each account goes

    Your ICP

    200+ staff, head office in the US or Canada, sells software to businesses, has a sales team, offers a free plan

    Act4

    yes, above the threshold

    Skip3

    no, above the threshold

    Review0

    below the threshold

      Enrich1

      unknown, so look it up

      HubSpot: yes, 0.65

      Meets 2 of 3; 1 can't be checked. Passes staff and head-office filters.

      Why you can trust each answer

      • They come from structured data

        We crawl the web ourselves and store what we find as typed fields with a source and a date. Decisions read those fields, so your agent never reads a raw page.

      • The confidence is calculated

        It reflects how much of what the question needs is in the record. Ask again and you get the same number.

      • Every answer is dated

        data_as_of and the tool version come back with every answer. With reasoning on, would_change_if lists what would change it.

      • Naming the check makes it repeatable

        In a pipeline, call is_icp_fit or is_contact_current by name and the same tool runs each time.

      34 questions Decisions can answer

      Each check below is one yes or no question. Your agent can ask in plain English and Firmable picks the right check, or your code can name the check it wants. A few, like the ideal customer check, need you to describe what you're looking for.

      Is this account a good fit?

      • Does it match our ideal customer profile?is_icp_fit
      • Does it belong in this campaign's audience?matches_campaign_audience
      • Is its headcount in the range we want?matches_company_size
      • Is it public, private, government or another type we want?matches_company_type
      • Does it sell to other businesses?is_b2b
      • Does it offer a particular service?offers_service
      • Anything else? Write your own yes or no question.evaluate_custom_decision

      What teams build with it

      Qualify accounts and find who to call

      1. Check each account against your ICP
      2. Find the decision makers at the ones that fit
      3. Keep only the people you can reach
      Your agent callsis_icp_fitis_decision_makeris_reachable

      Use it in your agent

      Paste one prompt and your agent connects itself, makes a test call and remembers when to use it. Then ask in plain English, or call the tool by name for a check that runs the same way every time.

      Set it up in one step

      Set up Firmable Decisions for me.
      
      1. Connect the Firmable MCP server by following
         https://agents.firmable.com/developer/setup.txt
         Keep any MCP servers that are already configured.
      2. If you need an API key instead of signing in, ask me for it. Keep it in an
         environment variable called FIRMABLE_API_KEY or our secret store, never in
         a file that gets committed.
      3. Check it works: call make_decision with tool is_publicly_listed and
         domain stripe.com, then show me the decision, its confidence and the
         reason.
      4. Add this rule to your project memory (CLAUDE.md, AGENTS.md or .cursor/rules)
         and tell me where you saved it:
         "For yes or no questions about a company or a person, call Firmable
          make_decision instead of searching the web. Quote the reason and
          source with each answer, and when the decision is unknown, ask me
          rather than guess."
      
      What Firmable Decisions can do: https://developer.firmable.ai/products/decisions.md
      Works in Claude, Cursor, Codex or any agent that can run commands.

      What the prompt does

      1. Connects the Firmable MCP server, by browser sign-in or your API key
      2. Keeps the key in an environment variable, out of your code
      3. Makes one test call and shows you the answer
      4. Saves a rule so your agent reaches for Firmable Decisions on its own

      Then ask, or call it by name

      Use Firmable to check whether each company in accounts.csv fits our ICP:
      50 to 500 staff, software, head office in the United States.
      Show the decision, the reason and the source for each.
      Works in any client with the Firmable MCP server.

      Every argument and field

      The docs list each argument, every field in the response and what each error means.

      Pricing

      $0.01per answer
      $0.02per answer with reasoning
      $25in free credit to start
      34checks, all the same price
      WhatCreditsUS$
      Yes or no answer1 credit per 25$0.01
      With reasoning2 credits per 25$0.02
      Unknown answers and errorsFree$0
      Listing the toolsFree$0

      Based on $0.25 a credit. Part-credits carry over between calls. Volume discounts apply. Full pricing

      Questions

      What is Firmable Decisions?

      An API and MCP tool that answers yes or no questions about companies and people from Firmable's own data. Each answer includes a confidence score and the facts behind it. With reasoning on, it also gives the source, the date and what would change it.

      How do I use it?

      Connect the Firmable MCP server to your agent and ask in plain English, or call POST /decisions or make_decision with a tool name, such as is_icp_fit, and the company's domain or Firmable ID.

      How much does it cost?

      One cent per answer, or two cents with the reasoning included. Unknown answers are free. New accounts start with $25 of free credit.

      Is unknown the same as no?

      No. Unknown means Firmable doesn't hold the fact needed to answer, so your agent knows to look it up before acting.

      What does confidence mean?

      How sure the answer is, from 0 to 1, worked out from the data we hold. Set a threshold and let your agent act above it.

      Why not use a model with web search?

      A web search answer depends on the pages it finds that day. Decisions answer from Firmable's own records, such as job changes and email deliverability grades, and every answer carries the date of its data.

      Can I ask my own question?

      Yes. Use evaluate_custom_decision to ask any yes or no question about a company.

      Where can I use it?

      Through the REST API (POST /decisions) or the Firmable MCP server, with a developer API key or by signing in. It works with Claude, ChatGPT, Cursor and other MCP clients.

      Are answers about people firm facts?

      No. Answers about people come back with advisory set to true: a guide, not a firm fact.

      Start with $25 in free credit

      Ask yes or no about a company. Get the answer and the facts behind it. No card needed.