Company data for AI agents

An AI agent that writes to a prospect, prepares a meeting or builds a list needs facts about companies. Without them it guesses, and a guess about a customer or a competitor looks bad in front of a buyer. The Innerfetch API gives the agent those facts, and each customer, competitor and vendor row comes with the public page that shows it. The agent can cite its source, and you can check it.

What an agent can do with it

• Research an account: industry, size, headquarters, tech stack, open roles and executives from one domain.
• Find who buys from a company: customers and partners, each with the page that names them.
• Map a market: the competitors of a company, and why each one is on the list.
• Build a list: companies that match a market you describe in plain words.
• Reach a person: the people of a company found on the public web, and a work e-mail check.

Option 1: the MCP server, with an API key

The MCP server gives your agent 12 tools. It uses the same key, prices and limits as the REST API. In Claude Code, add it with one command:

claude mcp add --transport http innerfetch https://api.innerfetch.com/mcp \
  --header "X-API-Key: $INNERFETCH_API_KEY"

The MCP server page has the setup for Cursor and VS Code. Also give the agent our skill file. It tells the agent which tool to use for a question, what each call costs and what each error means.

Tool	What the agent gets
company_details	The full company profile from a domain
company_customers	Customers and partners, each with its public page
company_competitors	Competitors, and the reason each one is on the list
company_vendors	Vendors that name the company as their customer
company_people	People of the company found on the public web
company_search	Companies that match a plain-words description
company_batch	Profiles for up to 50 domains in one call
work_email_lookup	A work e-mail and how sure we are
company_status, credits_balance, usage	Free checks on progress, balance and spend

A prompt that works

We gave an agent this prompt on 26 October 2026, with only the Innerfetch tools:

We sell a help desk for small support teams. Who are the main competitors of helpscout.com, and which companies are customers of those competitors? Give me 10 accounts to contact, each with the competitor it uses and the public page that shows it. Use the Innerfetch tools and tell me how many credits you used.

The agent checked its balance, asked for the competitors of helpscout.com, then asked for the customers of six of them, with a small max_results on each call. It returned a table of 10 accounts, each with the competitor and the URL of the competitor’s customer page. It told us which competitors had no public customer list, and how many credits it used.

Three things made the answer good: the prompt names a domain, it asks for the proof, and it asks for the cost. Add the same three to your own prompts.

Option 2: pay per call, with no account

An agent with a wallet can call the REST API with no key and no sign-up. A call without a key answers 402 with a quote. The agent signs the quote and sends the same call again, and pays in USDC for that one call. A queued call or an empty answer takes nothing. The pricing page shows the price per call, and llms.txt has the steps for the agent.

Use this for an agent that makes a few calls for many different users. Use an API key and the MCP server when one team uses the agent every day.

Option 3: your own agent code

Any framework that can make an HTTP call can use the REST API. Give the agent llms.txt or the full API reference, and one tool that calls GET /api/v1/company/details with a domain. Add the customers and competitors endpoints when the agent needs them.

Keep the spend under control

• Give the agent its own API key. You can revoke it in the console without stopping your other work.
• Tell it to set max_results or limit to the rows it needs. You pay for each row of a list.
• Tell it to poll company_status, which is free, while a new company is fetched.
• Read the usage tool or the console to see the credits by day and by call.