Introduction
LinkedIn actions for AI agents, across many accounts, with a safety layer the agent can't override.
SendCrew gives an AI agent a set of small LinkedIn tools: send_connection, send_message, check_inbox, reply_to_thread and more. Your agent decides who to contact and what to say. SendCrew runs the actions from your connected LinkedIn accounts ("senders") and keeps them safe.
What SendCrew does for you
- Queues every write. A write returns an
action_idstraight away and runs later in that sender's queue, one action at a time. - Enforces limits on the server. Daily and weekly caps, warmup for new accounts, quiet hours and pauses between actions. No parameter turns them off, so a looping agent can't burn an account.
- Never contacts the same person twice from two senders in one workspace.
- Stops on trouble. A checkpoint, captcha or restriction pauses the sender and tells you. Nothing retries through it.
- Reports back through
get_action_status, the unified inbox and signed webhooks.
Two ways in
MCP server
Claude Code, Cursor or any MCP client over Streamable HTTP.
REST API
Same operations for n8n, Clay, Make or your own scripts.
Both use one tool list, so every tool on the tools reference exists in both.
For agents
Machine-readable versions of these docs: /llms.txt and /api/v1/openapi.json.