LinkedIn tools for AI agents

Give your AI agent LinkedIn.

Read the feed and inbox, understand people, comment on posts, and run outreach — through MCP and ready-made agent skills.

Built-in LinkedIn limits pace every action automatically to reduce warnings, account restrictions, and ban risk.

MCP-native · Ready-made skills · Built-in LinkedIn limits

works with Grok · Claude · Cursor · OpenAI Agents SDK · any MCP client

agent · morning loop
01
Read feed
linbox_scrap_feed
02
Pick a relevant post
agent
03
Write a useful comment
linbox_comment_on_post
04
Published when limits allow
queued
comment queued · within limits
{
  "status": "queued",
  "task_id": "tsk_1a2b3c",
  "scheduled_for": "2026-09-03T09:20:00Z"
}

the agent keeps working. Linbox publishes when the account is still within safe activity levels.

[ 01 ]Running on my account

What I already run with Linbox.

These are not demo workflows. They run every day on my LinkedIn account.

Grok bot
A Grok bot that grows my visibility

Every morning, my Grok bot reads my LinkedIn feed. It finds posts where I can add something useful, chooses what to say, and publishes a relevant comment from my profile. I stay visible in the right conversations without scanning the feed all day.

  1. 01Read feed
  2. 02Rank relevant posts
  3. 03Write comment
  4. 04Publish

Consistent visibility, without a content calendar.

Grok bot
A Grok bot that turns engagement into outreach

The bot finds posts related to my product, loads the people who reacted or commented, and filters for relevant profiles. The best matches go into an outreach campaign that can connect, follow up, and stop when someone replies.

  1. 01Find relevant posts
  2. 02Parse reactions & comments
  3. 03Qualify people
  4. 04Enroll in campaign

Real buying signals become a prospect list — without CSVs or manual research.

Linbox gives an agent the full loop: observe → reason → act → follow up.

[ 02 ]How it works

Connect once. Let the agent work.

01
Connect your account

The Chrome extension captures the session. Linbox runs a server-side browser with a per-account proxy.

02
Add MCP and skills

Give your agent the MCP endpoint, then install the Linbox skills for tool guidance and ready-made playbooks.

03
Describe the outcome

Ask the agent to monitor the feed, manage the inbox, find people, or run an outreach workflow.

04
Linbox enforces the limits

Reads use fresh or cached context. Every action is checked against account limits and queued with a pollable task_id.

[ 03 ]Capabilities

The LinkedIn tools your agent actually needs.

Feed & posts

Read the home feed, search posts, and inspect content and engagement.

Comments & reactions

Publish comments, like posts, and load the people behind reactions and comment threads.

People & companies

Parse profiles and companies; search, tag, and filter the resulting relationship data.

Inbox & messages

List threads, read full history, reply, and detect new messages.

Connections

Send, accept, and withdraw invites with dedup and connection-state checks.

Outreach campaigns

Enroll people into multi-step flows that connect, message, wait for the agent, and stop on reply.

[ 04 ]Bring your own agent

Your agent already knows how to use Linbox.

MCP exposes the tools. Skills teach the agent the contract, the workflows, and what to do when work is queued.

MCP-native tools

Typed tools and structured JSON or markdown responses for Grok, Claude, Cursor, OpenAI Agents SDK, and custom agents.

Ready-made skills

Download skills for messaging, connections, tags, enrichment, and complete playbooks. Install all of them as a zip or pick only what your agent needs.

skill · linbox/SKILL.md
# when the agent needs LinkedIn
tools: linbox_scrap_feed, linbox_comment_on_post, linbox_enroll_campaign
Writes enqueue work. Poll linbox_get_task. Never retry a deferred action in a loop.
[ 05 ]Limits enforced automatically

Let the agent move fast — without letting it spam LinkedIn.

The agent does not need to understand or remember safe activity levels. Linbox checks every action before execution and enforces its own limits per minute, hour, day, week, and month, plus minimum intervals, work hours, and timezone. Configure them per account or start with a preset.

Your agent decides what to do. Linbox keeps it within the limits.

This cannot guarantee that LinkedIn will never restrict an account, but it materially reduces the risk of warnings, restrictions, and bans caused by aggressive automation.

Limits are enforced, not suggested

If the agent asks for too much, Linbox does not execute it immediately.

Per-action budgets

Reading the feed, sending messages, connecting, and commenting each have their own activity levels.

Work hours and timezone

Activity happens when a real person would be active.

Queue instead of failure

If an action should not run now, Linbox schedules it and returns task_id + scheduled_for.

queued · scheduled_for 14:20

The agent gets a task_id, moves on, and picks up the result later. No dead ends.

[ 06 ]Get started

What could your agent do with LinkedIn?

Connect your account, add the MCP, and start with a ready-made skill.