Claude connector
Run your LinkedIn outreach from Claude.
Gigi reads each prospect's profile and company page, then writes a message that could only have been sent to that one person. Nothing goes out until you approve it. Connect it to Claude and the whole loop lives in your chat: approve drafts, answer replies, work your inbox, retarget campaigns, pull the numbers.
Who replied this week, who is waiting on me, and what does that say about my targeting?
crm_pipeline_reportGigi: list_replies118 requests sent, 55 accepted, 34 replied. Six people are waiting on you. The people who engaged look nothing like the people who ignored you.
Founders at 10 to 40 person studios, Nordics and Benelux. Nineteen of the 22 mentioned hiring or capacity.
avg fit 88 · 4 marked hot
Managing directors at 100+ person consultancies. Longer approval chains, no hiring signal on the company page.
avg fit 71
Suggested change
Raise the fit threshold to 80, narrow headcount to 10 to 50, drop “consultancy” from the keywords. Remaining pool: 340 prospects.
Want me to apply that to the campaign, and draft replies for the six who are waiting?
One question, the whole readout: numbers, patterns and the change Gigi would make, waiting on your yes.
Why from Claude
Your Claude already knows your business. Gigi campaigns start there.
Every other outreach tool starts you on an empty form: paste a job title, pick a country, hope. Claude has been in your context for months. The positioning you rewrote, the deals that closed, the objections that killed the rest, the docs and transcripts already in your project. Ask for a campaign there and the targeting and the copy inherit all of it.
Gigi then does the per-prospect work: profile, company page, recent activity. Your context sets the strategy, Gigi's research writes the line.
- Your positioning doc and pricing pageMessages that lead with the value prop you actually sell on
- The deals that closed last quarterTargeting drawn from who really buys, not a job-title guess
- Call notes and the objections you keep hearingFollow-ups that answer the objection before it gets raised
- Last month of replies and fit scoresThresholds, keywords and sending pace tuned to what worked
No re-briefing, no exporting notes into another dashboard. The context is the brief.
Setup
3 steps, two minutes to complete. Then just chat.
Use the account you already have
Sign in at usegigi.ai with LinkedIn. The connector runs on that same account: same campaigns, same credits.
Nothing separate to create
Paste the connector URL
In Claude desktop or claude.ai: Settings, Connectors, Add custom connector. Name it Gigi, paste the URL.
About 40 seconds
Allow once, then just ask
Claude opens a Gigi page, you read what the connector can do and click Allow. Then type “list my campaigns”.
Standard OAuth, revoke any time
Connector URL
https://usegigi.ai/api/mcpSettings, Connectors, Add custom connector. Name it Gigi, paste the URL, then Allow on the Gigi page that opens.
Using Cowork?
Add the Gigi skill too. Claude then follows the review-before-approve workflow on its own, without you restating the rules each session.
Get the Gigi skill fileWhat it does
One less interface to learn.
You already spend the day in Claude. So there is no new dashboard to log into, no tab to remember to check. Ask for the draft queue, ask what came in overnight, ask why a campaign slowed down. Every job below runs in that same thread, and the Gigi app stays exactly where it was for the days you would rather click.
The one people use daily
Review and approve drafts
Ask for the queue and Claude reads back the connection requests and follow-ups Gigi wrote, with the fit score and the hook behind each one. Rewrite the weak lines in plain language, approve the rest in a batch.
- BatchApprove outreach drafts together
- One by oneReply drafts, after you read the reply
- NeverA send without a human behind it
Triage replies by sentiment
Replies grouped positive, neutral, negative, each with the answer Gigi drafted. You read the reply, then decide what goes out.
Work your inbox
Inbox CRMEvery LinkedIn conversation with its stage, tags, notes and follow-up date. Ask who is waiting on you, move a contact to hot, set a follow-up, have Gigi draft the reply. Power-user accounts.
Launch and steer campaigns
Build one from an ideal-customer description or a Sales Navigator search. Change target, keywords, fit threshold or sending window mid-flight. Pause, resume.
Reports and exports
Weekly numbers per campaign, fit-score histograms, CSV exports, and Copilot questions at the same credit cost as the app.
Credits and invoices
Check your balance, pull a Stripe Checkout link for a pack, list invoices, without leaving the conversation.
How it works
Ask the way you would ask a colleague.
There is no syntax, no command list and no field names to learn. You describe what you want, Claude works out which part of your Gigi account that touches, and the answer comes back from your live data rather than from the model's imagination.
Anything that only reads is instant. Anything that changes or sends stops and waits for you.
- 01
You ask in plain words
A question, a complaint, half an idea. Claude reads it against what it already knows about your business and picks the Gigi action that fits.
- 02
Gigi answers from your account
Counts, fit scores, drafts, replies and credits are fetched live from your own campaigns. Nothing in the answer is invented.
- 03
Claude comes back with a proposal
The readout, the drafts it would send, or the change it would make to a campaign, written out in the thread with the reasoning beside it.
- 04
You approve, then Gigi acts
Only after your go-ahead does anything move, and it moves through the same paced sender, daily cap and sending window as the app.
Guardrails
Claude does the legwork. You keep the veto.
The connector is another door into your own account, not a new set of powers. Every limit that keeps Gigi accounts off LinkedIn's radar applies to anything Claude does.
- Human approval on every send. You read the drafts, you say which ones go. Claude cannot skip the approval, the daily cap or your sending window.
- Your account only. The connector runs as you, with the app’s permissions. It never sees another user’s data.
- Standard OAuth, no keys to paste. Revoke from Account, Connected apps. The token dies immediately, on both sides.
- Everything is audited. Each change the connector makes is written to the campaign timeline, labelled as a connector action.
FAQ
Questions, answered.
The connector is free. Installing it, connecting your account and asking it questions costs nothing on top of your Gigi plan, and there is no surcharge for working through Claude. What you pay for is the work itself, metered exactly as it is in the app and drawn from the same credit balance: finding and researching a prospect, sending an approved message, asking Gigi Copilot, drafting an inbox reply. Reading campaigns, drafts, replies, reports and invoices is free in both places. A campaign run from Claude costs the same as the identical campaign run from the Gigi interface, to the credit.
No. Gigi never sends a message a person has not approved, in the app or from Claude. The connector approves what you tell it to approve, and sends run through the same paced, capped sender. Inbox reply drafts wait in the inbox until you send them yourself.
Only your own Gigi account: your campaigns, prospects, drafts, replies, inbox, reports and credits. It cannot reach other accounts, your LinkedIn login, or admin functions. Managed accounts get read and review tools only.
Account, Connected apps, Disconnect. The token stops working immediately. You can also remove the connector on the Claude side.
Claude desktop, claude.ai on plans with custom connectors, Cowork, and Claude Code. Anything that supports remote MCP servers with OAuth.
Free to connect. 40 credits to start.
Paste one URL into Claude and your outreach lives where you already work. Credits only burn on the same actions they cost in the app.
Connector URL
https://usegigi.ai/api/mcpUsing Cursor, Windsurf or another MCP client? Gigi is also on Smithery.