UseFireApp
PricingSign inStart free trial

← Product

AI agents

Run your outreach from the AI agent you already use

Ask Claude, Cursor, or any agent that speaks MCP to build tomorrow's list, explain a quiet week, or work through this morning's approvals. It does the work in your workspace, under the same limits you have.

What you can hand it

  • Find the right peopleDescribe who you want to reach. It runs the search and shows you who it found, and nobody is added until you choose. Once they're in, it scores each one against your targeting.
  • Fill a sequenceIt imports the people you pick, tags them, and starts them on a sequence. Anyone who can't start comes back with the reason.
  • Reach outA connection request or a message to one person, with words you wrote together.
  • Work the inboxIt lists the threads waiting on you and drafts replies for you to edit.
  • Clear approvalsGo through the approval queue together. Edit a message before it goes, approve the rest in one pass.
  • Explain what happenedWhy someone is on your list, what they've been sent, and how the last week or month went, down to the names behind each number.
  • Keep people off the listAdd a person, a company, or a domain to your do-not-contact list, or archive someone and stop everything waiting to go to them.

Things to ask it

When something looks off
  • “Why did it only send 12 today?”
  • “Why is this person on my list, and what have we sent them? linkedin.com/in/…”
  • “How did the last 30 days go? Who replied?”
  • “Which threads have been waiting on me the longest?”
When you want something built
  • “Find finance directors at logistics companies in Texas and show me who you found.”
  • “Import the ones I marked, score them against my targeting, and start the strong fits on my standard sequence.”
  • “Add every company on this list to my do-not-contact list.”
When the day starts
  • “What's waiting on me today? Approvals first, then threads that need a reply.”
  • “Draft replies to everyone who asked a question this week. Don't send anything.”
  • “Approve everything in the queue except anything going to a current client.”

What it won't do without you

The agent acts with a key you create, inside your limits and your do-not-contact list.

  • Get in without your keyYou create the key in Settings → Integrations and can revoke it at any time. A key reaches the one workspace it was made in.
  • Send past your limitsIts connection requests and messages count toward the same daily and weekly limits as yours. At the limit, it's told no.
  • Reach anyone you've ruled outPeople, companies and domains on your do-not-contact list can't be contacted from an agent either.
  • Send anything you can't see laterEvery connection request and message it sends shows in that person's Activity, next to the ones you sent.
  • Go around your approvalsIf your workspace holds sends for approval, the agent's sends wait there for you. Approve them yourself or ask the agent to. Without approval holds, its sends go out right away, within your limits.
  • Change your planIt can read your plan and usage. Upgrades happen in Settings, by you.

Set it up

  1. Create a key

    Settings → Integrations → Create key. Copy it right away; it's only shown once.

  2. Add it to your agent

    Paste your server address and key into Claude Code, Cursor, Claude Desktop, or any MCP client.

    Setup by client

  3. Give it the playbook

    Optional. Add the agent skill file so it knows the usual workflows.

    Agent skill file

  4. Ask one question

    Start with something that only reads, like what's waiting on you today.

  5. Automate the rest

    The same key works with the REST API, for scripts or tools like n8n and Make.

    API reference

Questions people ask first

Do I need to write code?
No. If you can paste a key into your agent's settings, you can connect it.
Does it get around LinkedIn's limits?
No. It sends from your connected LinkedIn account, under the same daily and weekly limits as the app.
How is this different from the API?
Same key, same rules. A script runs the steps you wrote. An agent picks the steps from what you ask it in plain language.
Can I connect more than one agent?
Yes. Give each one its own key so you can revoke them separately.
Can it see anyone else's workspace?
No. A key only reaches the workspace it was created in.

Start with one question

Connect your agent and ask what's waiting on you today. Nothing sends from a question.

Start free trialRead the setup guide