150 accounts and 300 people, 300 LinkedIn connection requests and a message to the 90 or so who accept.
Give this to your agent
Using RouterGrowth, run this workflow: Connect on LinkedIn with founders who posted about AI agents for sales or marketing this month, then message the ones who accept. 300 people, personalized from their posts. Show me the price before every call, and stop after each stage with a sample of the results so I can adjust. 1. Source the accounts: company.search x 15 searches. Companies matching the industry, size and location, ten or so per search. 2. Qualify on firmographics: company.enrich x 150 companies. Headcount, industry and location to keep the accounts that fit. 3. Find the people: linkedin.search x 150 searches. People search on your connected LinkedIn account, one search per account. 4. Get the contact: linkedin.profile x 300 profiles read. Charged on a match only: a miss is free, so the estimate is a ceiling. 5. Context for the first line: web.scrape x 150 pages, social.posts x 900 posts. One page per account and a few recent posts per person, so the opener is about them. 6. Run the outreach: linkedin.invite x 300 requests, linkedin.message x 90 (30% accepted). Requests and follow-ups from your connected account, inside LinkedIn's daily limits. 7. Read the replies: linkedin.messages x 30 inbox reads. Poll the inboxes, classify replies, hand the warm ones to a human. Budget for one run: about $8.11. Run routergrowth inspect on any capability before the first call to confirm the quote.
Works in Claude, ChatGPT, Claude Code, Codex and Cursor. New workspaces start with $1 of credit. Get a key or read the setup guide.
Why it works
Founders posting about AI agents are early adopters by definition and reply to messages about the thing they just wrote.
On LinkedIn the signal and the channel are the same place: what someone posted is the reason to connect, and the connection is the conversation. These workflows find people by role and segment, read their recent posts, send a connection request with a note about what they wrote and message the ones who accept.
The first line, as your agent would write it
Your post on running outbound with agents matches what we see. Which step still needs a human for you?
The workflow
7 stages and 2,085 calls on one key. Every call is quoted and reserved before it runs, and a lookup that finds nothing releases the hold.
Companies matching the industry, size and location, ten or so per search.
Headcount, industry and location to keep the accounts that fit.
People search on your connected LinkedIn account, one search per account.
Charged on a match only: a miss is free, so the estimate is a ceiling.
One page per account and a few recent posts per person, so the opener is about them.
Requests and follow-ups from your connected account, inside LinkedIn's daily limits.
Poll the inboxes, classify replies, hand the warm ones to a human.
The receipt
About $8.11 for the whole run. No subscription and no seats: each call is paid from a prepaid balance at the price shown before it runs.
| Capability | Calls | Cost |
|---|---|---|
| company.search$0.0024 per search | 15 | $0.04 |
| company.enrich$0.0071 per match | 150 | $1.06 |
| linkedin.search$0.002 per search | 150 | $0.30 |
| linkedin.profile$0.002 per profile | 300 | $0.60 |
| web.scrape$0.0056 per page | 150 | $0.84 |
| social.posts$0.002 per result | 900 | $1.80 |
| linkedin.invite$0.0088 per invitation | 300 | $2.63 |
| linkedin.message$0.0088 per message | 90 | $0.79 |
| linkedin.messages$0.002 per read | 30 | $0.06 |
| One run | 2,085 | $8.11 |
Assumptions: 2 contacts per account unless the brief says otherwise, a 70% email find rate, 90% of found addresses passing verification, 3 emails per contact and a 30% LinkedIn accept rate. Prices are the catalog's list prices when this page was built. Change the volumes and the assumptions in the workflow generator, or see how plans lower the price of every call.
# RouterGrowth workflow: Connect on LinkedIn with founders who posted about AI agents for sales or marketing this m
npm install -g routergrowth && routergrowth setup
# 01 Source the accounts (company.search x 15)
routergrowth run -c company.search -i '{"query":"<query>"}'
# 02 Qualify on firmographics (company.enrich x 150)
routergrowth run -c company.enrich -i '{"domain":"<domain>"}'
# 03 Find the people (linkedin.search x 150)
routergrowth run -c linkedin.search -i '{}'
# 04 Get the contact (linkedin.profile x 300)
routergrowth run -c linkedin.profile -i '{}'
# 05 Context for the first line (web.scrape x 150, social.posts x 900)
routergrowth run -c web.scrape -i '{"url":"<url>"}'
routergrowth run -c social.posts -i '{"platform":"<platform>","handle":"<handle>"}'
# 06 Run the outreach (linkedin.invite x 300, linkedin.message x 90)
routergrowth run -c linkedin.invite -i '{}'
routergrowth run -c linkedin.message -i '{"text":"<text>"}'
# 07 Read the replies (linkedin.messages x 30)
routergrowth run -c linkedin.messages -i '{}'
FAQ
Yes, from your own account. You connect it once through a hosted sign-in, then linkedin.search, linkedin.invite and linkedin.message run as you, inside daily caps the platform enforces.
Up to 30 a day per account, 20 in the first week, and 180 a week. The limit is enforced on every call, so a workflow of 300 people is spread over a couple of weeks.
One line about something the person posted. social.posts reads their recent posts so the agent can write a note that could only have been sent to them.
linkedin.messages shows who accepted and who replied. The agent messages the people who accepted and, in the workflows that say so, emails the ones who did not.
About $8.11 for one run of 2,085 calls. Each call is quoted before it runs and a lookup that finds nothing is not charged, so the figure is a ceiling. There is no subscription: the run is paid from a prepaid balance.
Keep going
Reading this as an agent? This workflow as markdown: /workflows/linkedin-ai-agents.md · every workflow: /workflows.md
Paste the prompt into your agent. The first calls run on the $1 that comes with signup.