Claude Code plugin for LinkedIn outreachWarmline
Warm outreach
that sounds like you.
Warmline reads your LinkedIn, Instagram and Skool inboxes through Claude in Chrome, finds who is waiting on you and who is due a follow-up, and drafts every message in your voice. Nothing is sent until you approve the exact text.
Mostofyourclientscomefromconversationswithpeoplewhoalreadyknowyou.Warmlinemakessurenoneofthoseconversationsgocold.
Every day, in one command
Say “do today’s outreach.”
Warmline runs one desk per account in parallel, each in its own Chrome profile, then puts the whole day on one review page.
Never miss a reply
The people waiting on you come first.
Every run reads every thread where the lead wrote last, including LinkedIn's Other tab and message requests. They stay at the top, oldest first, with their last message word for word, until you reply.

Evidence on every card
Every fact checked in your browser first.
Before a draft mentions anything about a person, Warmline opens their profile in Chrome and saves a screenshot of it on the card. If it is unsure of a fact, the message does not go.

Your voice
Drafts that read like you wrote them.
Paste a few messages you sent yourself. Warmline learns your greetings, your length and your habits, keeps learning from every draft you rewrite, and runs each one past a verifier that reads it as the person receiving it.

One page to decide
Accept, edit or reject. One card at a time.
A local review page shows each person with the playbook step, the evidence and the draft. Edit inline, accept or reject, and every click saves the moment you make it.


The approval guarantee
Nothing sends until you say so.
Enforced in code at every step between your yes and the send.
Exact text, hashed
Each approved message is bound to a SHA-256 of its exact text, and the batch to a hash of the whole table.
Your yes, captured by a hook
A hook records your confirm phrase only when you type it in chat. The model has no way to write that record.
Re-checked before every send
Right before typing, each sender checks the item is still approved with the same hash. Edited means not sent.
Chrome typing is guarded
A second hook refuses to type any draft into the browser that was not approved with the same hash.

And everything around it
A whole outreach desk, on your laptop.
Your whole pipeline
Stage, playbook step, last contact and nudges for every lead, with several niches per account.

Today in numbers
Reached, replied, interested, calls and won, for 7, 30 or 90 days.

What only you do
Replies in live conversations are drafts for you to send yourself.

Campaign boards
Collab invite campaigns move people through stages as things happen.

Parallel desks
One agent per account, each in its own Chrome profile and tab, all at once.
Paced, careful senders
About one message a minute. Any warning, CAPTCHA or login challenge stops that platform for the day.
Playbook presets
Five-step warm outreach, generic B2B, or coach and creator. Edit playbook.md freely.
Filters and greeting rules
Blocked countries and industries, words you never use, formal greetings only when the profile shows it.
Weekly review
/warmline:weekly sums up what worked and puts this week's focus on the Today page.
A doctor for setup
/warmline:doctor checks everything and prints the exact fix for anything that fails.
Works where your conversations already are.
Private by design
Your data stays on your machine.
Contacts, messages and screenshots live in a workspace folder you choose. The review page listens on 127.0.0.1 only. Claude Code still talks to Anthropic's API as in any session.
Up and running in minutes
Three steps. Then just ask.
- Step 1
Add the plugin
Unzip the download, then run these in Claude Code and restart when it asks.
> /plugin marketplace add ~/Downloads/warmline> /plugin install warmline@warmline
- Step 2
Create your workspace
Answer a few questions about you, your offer and your accounts. Add --demo to try it with made-up leads first.
> /warmline:init ~/warm-outreach> /warmline:doctor
- Step 3
Run your day
Open the review page it links, decide each card, and type your confirm phrase.
> do today's outreach
Tech specs
- Works with
- Claude Code with plugin support, and the Claude in Chrome extension
- Platforms
- LinkedIn, Instagram and Skool. Reddit and Facebook optional
- Accounts
- Several per platform, each in its own Chrome profile
- Runs on
- macOS or Linux. Windows through WSL2
- Requires
- Python 3.9+ (standard library only). OpenSSL 1.1.1+ or age for backups
- Data
- Local SQLite database in a folder you choose. No server, no telemetry
- Review app
- Local web page on 127.0.0.1, already built. Optional token-protected LAN mode
- Licence
- Single seat: one named person, for their own business outreach
- Updates
- Included. Your download link always serves the latest version
- Current version
- v0.6.2
Questions, answered.
No tool can promise that. Automated activity may breach the terms of LinkedIn and other platforms, which can lead to warnings, limits or a ban. Warmline keeps the risk low: small daily caps, 10 or more seconds between profile views, about a minute between messages, no likes or endorsements, and an immediate stop at any warning, CAPTCHA or login challenge. It contains nothing to hide automation. Read each platform's current terms and decide for yourself.
No. Every message, connection request, follow and comment is shown with its exact text and sent only after you type your confirm phrase for that table. If you edit a message after confirming, it is shown again for a new confirmation. Replies inside live conversations are drafts that you send yourself.
A plugin: it bundles skills, slash commands, agents, hooks and a local review web app. You install it once with /plugin and it adds the /warmline commands to Claude Code.
The desks and senders need Claude in Chrome, because they read and send through your own logged-in browser. Without it you can still use the review page, analytics, CSV import, backups and the weekly review.
Not natively. Use WSL2, with Chrome and Claude in Chrome on the Windows side. macOS and Linux are supported directly.
In the workspace folder you choose, on your machine. Warmline has no server, no account, no licence check that phones home and no telemetry. Claude Code sends your conversation to Anthropic's API as it does in any session, including what the agents read in your browser.
Drafts follow a voice profile learned from messages you wrote, including your greetings, length and grammar habits. Each draft goes through a lint for your rules, a humanizer pass and a verifier agent that reads it as the recipient. You still read and approve every one.
Yes. /warmline:init <folder> --demo creates a workspace with synthetic leads and a ready day, so you can explore the review page safely.
No. The licence is one seat for one named person, for their own business outreach. Offering it as a service to others is not allowed.
Your download link always serves the latest release. An update replaces the plugin folder and leaves your workspace as it was, and every release is listed in CHANGELOG.md.
Your warmest leads are already in your inbox.
Warmline makes sure you answer every one of them.