in LinkedIn API Live in production. Email, messengers & calendars next
Infrastructure
Messaging
Scraping
Enrichment data
Documentation Browse all docs →
QuickstartOne key, first call in 2 min MCP serverTyped tools for any agent API referenceEvery endpoint & schema WebhooksEvents, retries, signatures
GTM engineersConnect any agent to your GTM process AI SDRRun an autonomous SDR on real infrastructure AgenciesOnboard clients safely, run from your stack PlatformsAdd LinkedIn on proven infrastructure LinkedIn automationSafe connects, messages & follow-ups Account providersWarm, secure & lease account fleets
Pricing Blog
LinkedIn shadowban checkerDo your invitations still arrive? A ten-minute check All free toolsSmall checks that run without an account
Get API key Sign in

Use case · AI SDR

Your AI SDR, on one API.

The agent researches accounts, writes personalized first lines and runs the outreach itself, through typed MCP tools on your own LinkedIn accounts. Every send stays inside per-account safety limits it cannot override.

Get your API key
People searchSales Navigator · 1,204 results
Example LinkedIn profile photo Maria LopezFounder & CEO · Brightleaf Agency agency Connect
Example LinkedIn profile photo David KimCMO · NorthPeak Marketing C-level Connect
Example LinkedIn profile photo Sofia RossiFounder · Rossi Creative agency Connect
Example LinkedIn profile photo Raj PatelHead of Growth · Flowstate new Connect
The agent researches its own pipeline: search, returned as typed data

The full SDR loop, not a draft assistant

Research

People and company search plus enrichment give the agent real, current context before it spends a touch.

Write and send

The opener is grounded in what the prospect actually did, and the send happens through the same contract, no copy-paste step.

Follow up

Replies land as webhooks. The agent reads the thread, drafts the answer and keeps the sequence moving.

Agent-native by design

Real tools, not browser puppeteering.

Claude, ChatGPT, Cursor or your own runtime: add the MCP server with one key and the LinkedIn tools appear, typed end to end:

  • Preview-then-confirm on every send, so the agent shows its work
  • Fire-on-success: a row is written only when LinkedIn confirms
  • The same schema over REST for your backend and cron jobs
# the SDR loop as the agent sees it, typed MCP tools
scrape_linkedin_search_people  # build the list
enrich_linkedin_person_full_profile       # enrich before writing
send_linkedin_connection_request          # preview, confirm, send
send_linkedin_message                     # the follow-up thread

# or the same actions over REST, one schema
POST /api/linkedin-scraping/search-people
POST /api/linkedin-enrichment/person-full-profile
POST /api/linkedin-messages/send
The same typed contract for MCP and REST

Timing is the personalization

Reach out when something actually happened.

An SDR that opens with last week's news loses the reply. The agent watches the accounts it works and moves on the change:

  • New posts, fresh engagers and inbound invites, surfaced as data
  • Enrich and qualify before the first touch, not after
  • Webhooks fire the connect or the message the moment data lands
Watchlistauto-scraped
Keyword comment Example LinkedIn profile photoAna Ruiz commented "PLAYBOOK" on your post DM the link
New engager Example LinkedIn profile photoErik P. reacted to your CEO's post Send connect
Event attendee Example LinkedIn profile photoNora Ø. attending SaaStr 2026 Personalize note
Scheduled re-scrapes of the people and companies that matter

The part that keeps it running

Autonomy inside guardrails.

The reason most AI SDRs die is not bad copy, it is a banned account. Here the safety layer is server-side, so the agent cannot outrun it:

  • Warm-up curves and health-aware daily limits per account, enforced before every action
  • Each account runs in its own cloud browser with a dedicated residential proxy
  • An activity log of every action, so you can audit what the agent did and when

GTM API reports 20,000+ accounts in production at under a 1% ban rate on this stack.

LinkedIn accounts2,418 connectedscroll →
AccountStatusCloud browserProxyGeoDaily limitAcceptanceHealthLast active
Example LinkedIn profile photoEmily Westwood@emily-westwood Running Chrome 124 res · DE DE 14/20 61% 96 2m
Example LinkedIn profile photoJordan Reyes@jordan-reyes Running Chrome 124 res · US US 9/20 44% 88 5m
Example LinkedIn profile photoAnya Müller@anya-mueller Running issue Chrome 123 res · DE DE 6/20 38% 72 18m
Example LinkedIn profile photoDaniel Kim@dkim-revenue Login issue Chrome 124 res · GB GB 0/20 - 41 3h
Example LinkedIn profile photoSofia Bianchi@sofia-b Running Chrome 124 res · IT IT 18/20 57% 94 1m
Example LinkedIn profile photoMarcus Hale@marcus-hale Running Chrome 124 res · US US 11/20 49% 90 8m
Every account the agent uses: own browser, proxy, budget, health

The stack under your SDR

20,000+
accounts run in production
<1%
ban rate
160+
typed LinkedIn tools
minutes
from key to running agent

Production numbers self-reported by GTM API.

Ship your AI SDR this week

One key into Claude or ChatGPT, connect an account, and the agent runs real outbound today. Start on the free plan, no card.

Start free