Floom

Example output

Scored outreach plan

3 interactions reviewed, 1 strong match

Anna Keller, Head of Partnerships, Cint

Relevance: 92 / 100. Runs partnerships at a survey and market-research platform, the exact profile in your brief, and viewed your profile twice this week. Warm and on-target.

**Connection note (draft, 214 chars):**

Hi Anna, saw you stopped by my profile. We run a consumer panel that plugs respondents into MR platforms like Cint; feels like there is an obvious partnership here. Would love to compare notes.

**First message (draft):**

Thanks for connecting, Anna. Quick version: we can supply vetted panel respondents into Cint's demand at scale. Worth a 20-minute call to see if the fit is real?

Draft mode: nothing was sent. Approve, edit, or reject each action.

What this delivers

Turn inbound LinkedIn attention (profile views, post engagement) into ranked, warm outreach.
Draft partnership connection notes that reference why each person actually fits your product.
Keep a human approval gate on every message while you scale outreach.

Overview

A partnership-outreach copilot for founders, modeled on the "engagement agent" pattern (who-viewed-you / post-reactors / new connections -> relevance score -> personalized note). It does three things:

  • •MONITOR — reads recent LinkedIn interactions (profile views, post likes &

comments, new connection requests). Out of the box you paste the interaction list; connect a LinkedIn account to pull them live.

  • •SCORE — ranks each person 0-100 for how relevant they are to YOUR goals,

grounded only in the product/partnership brief you provide (no guessing).

  • •DRAFT — for the relevant ones, writes a LinkedIn connection note (<300 chars)

and a short first message, personalized to why THIS person fits.

SAFETY MODEL (this is the point):

  • •Approval-gated by default. The worker proposes the outreach and STOPS —

it emits a decision for you to approve, edit, or reject. Nothing is sent on the first run.

  • •Per-action scope. It only ever reads interactions and drafts/sends outreach —

it cannot touch anything else in your account.

  • •No scraping, no fake browser session. Live monitoring + sending run through

a LinkedIn account you connect yourself , via Floom's managed LinkedIn connection (an account-pinned proxy; the worker never holds the vendor key), on your own authenticated account, with your consent — not by scraping linkedin.com.

  • •Every run is logged and reviewable.

Turn the approval gate off per action only when you want autonomous send; the default is human-in-the-loop.

What you provide

product_contexttextrequired

Paste what you do and the kind of partner you want. This is the ONLY grounding for scoring — the worker won't invent context.

interactionstext

Paste who engaged (profile views, post likes/comments, new connections). Leave blank once a LinkedIn account is connected to pull them live.

modeselect

draft = drafts only, you send. approve = propose + wait for your approval, then send (default). auto = send without asking (only when a LinkedIn account is connected).

How it works

  1. 1read interactions (pasted, or live via your connected LinkedIn account)
  2. 2score each person 0-100 vs your product/partnership brief
  3. 3draft connection note (<300 chars) + first message for the relevant ones
  4. 4emit an approval decision (draft mode: STOP here; nothing sent)
  5. 5on approval, send via your connected LinkedIn account (or hand back ready-to-send text if not connected)

Under the hood

The worker's instructions and configuration

Install with an agent

Give this to your coding agent. It sets you up first via the onboarding prompt, then imports this template and keeps the first run in preview mode.

Set up Floom in my coding agent, then add this worker. First, read https://floom.dev/onboard and follow it to install Floom and sign me in. Import this Floom worker template into my workspace and keep it in test mode first:
Add to workspace
Partnership Signal & Outreach | Floom