My Product

UseFireApp: LinkedIn and Email Outreach That Runs Every Working Day

My own outreach product. Sequences run LinkedIn and email every working day, stop the moment someone replies, and never send faster than a careful person would.

usefireapp.com
The UseFireApp homepage, showing a first week of outreach capped at ten connections a day
01

Problem

Outreach is the first thing to slip when a week gets busy, and the cost shows up weeks later as an empty calendar. Tools that automate it tend to send as fast as they can, which is how LinkedIn accounts get restricted. Recruiters, founders and coaches needed outreach that runs on schedule without putting the account at risk.

02

Approach

  • Built sequences that run LinkedIn and email every working day inside a sending window, and stop as soon as a prospect replies.
  • Made the send limits structural. New accounts start at 10 connection requests a day and warm up to 20, with weekly ceilings and a one-switch halt.
  • Gave AI agents a proper way in: an MCP server for Claude Code, Cursor and similar tools. A key only sees its own workspace, and its writes share the same rate limit as a person clicking through the app.
  • Added an approval queue so nothing sends without sign-off, and Spark Autopilot to plan each day's outreach for review.
  • Built white-label and coach access in from the start: a deployment under someone else's name, colours and domain, with operator access scoped to one workspace.
03

Outcome

  • A live product at usefireapp.com with public Solo and Business plans and a 7-day trial.
  • The same safety rules apply whether a person, an API key or an AI agent is doing the sending.
  • Sign-in by email code or passkey, running on Cloudflare's edge platform.

Stack

Next.js and React on Cloudflare WorkersCloudflare D1, Queues and Durable ObjectsWorkers AI for draftingMCP server for AI agentsStripe billing and WebAuthn passkeys

Putting an AI agent in front of real users?

Tell me what it needs to do. I'll tell you where the guardrails go.
Last updated: