Outbound for the agent era

Give your agents hands on LinkedIn. Keep the final say.

Connect your agent over MCP or the API and point it at your leads. It runs the outreach and drafts every reply inside guardrails built for LinkedIn. Your team reads, edits and sends, managing every lead from one inbox.

Read the docs
  • Sandbox free forever, MCP and the API included
  • Guardrails on every action

Screenshot
The full workspace: the inbox, a thread, and an agent draft waiting on a person

  • Logo
  • Logo
  • Logo
  • Logo
  • Logo
  • Logo

Teams running LinkedIn outreach through their own agents

How it works

From your agent to your inbox in three steps.

  1. 1

    Connect

    Point any agent at Sprootal over MCP or the REST API. An LLM with MCP support is enough to start, with no orchestration to stand up first. Minutes, not weeks.

  2. 2

    Act

    The agent gets the full LinkedIn action set: profile visits, connection requests, messages, sequences and reply drafts. Every action passes through Sprootal’s limits before it reaches LinkedIn.

  3. 3

    Decide

    Every live conversation lands in one inbox. Your team reads, edits, sends, and carries the dialogue forward.

Guardrails

Agents move fast. Sprootal keeps them inside the lines.

Limits are enforced on the action, not suggested in a setting. An agent in a loop reaches the cap and waits.

NNconnection requests per sender each day
NNmessages per sender each day
NNday warm-up ramp for a new sender
1dedicated proxy per sender, never a shared pool
  • Per-sender caps, counted per action type rather than in one bucket
  • Action-level enforcement: the limit is checked when the call arrives, so an agent cannot spend a week of allowance in an afternoon
  • Working hours you set, with the day’s work paced across the window
  • A person on send by default, which is the last control and the one that matters most
The numbers behind the guardrails

Every LinkedIn action, as an endpoint

Prospecting, connecting, messaging, sequencing and reply drafting, all callable by the agent you already run. A native MCP server for Claude, Cursor or n8n, a REST API for everything else, and sandbox access on the free tier.

More on this

Screenshot
Claude connected over MCP, calling the action set

Your agent opens the conversation. You carry it.

Drafted replies wait for review, with the agent and the playbook that produced them named on each one. Edit in place, send from one inbox across every sender, and keep the thread moving.

You can delegate sending to the agent for any conversation. Most teams reserve it for low-stakes follow-ups.

More on this

Screenshot
The inbox, with an agent draft waiting on a person

Wire it into the rest of your stack

Webhooks fire on every reply, every draft and every stage change, so your agent wakes the moment there is something to answer. Scoped keys per integration keep one agent’s actions separable from another’s.

More on this

Screenshot
Webhooks and scoped keys, one per integration

What customers say

I was spending 3 hours a day on LinkedIn DMs. Now the co-pilot drafts the replies and I spend that time closing.

Jeffrey HamelinkCEO at SmallGiants

From 40 hours a month on LinkedIn to 5, reviewing what the agent did. The best return of anything we have bought for sales.

Anastasiia SukhorukovaCEO at Foxi Education

Sprootal runs outreach for NN teams, from solo founders to agencies with a hundred senders.

Customer stories

Questions we get asked first.

Will this get my LinkedIn account banned?

Nobody can promise zero risk on an account they do not own, so here is the mechanism instead. Every action is rate limited per sender and counted per action type, not in one shared bucket. The limit is enforced when the call arrives, so an agent in a loop waits rather than spending a week of allowance in an afternoon. New senders ramp through a warm-up before they reach full volume, each sender keeps its own dedicated proxy, and work is paced across the hours you set. By default a person approves every send, which is the last control and the one that matters most.

Can the agent send replies automatically?

Yes, if you allow it, per conversation or per campaign. The default keeps sending with your team because a reply is where the deal is either carried forward or lost, and that judgment is worth a person’s attention.

Do I need to build an agent first?

No. Any LLM with MCP support plus Sprootal is enough to start. You point it at the MCP server with a scoped key and it has the full action set from the first call.

Which agents work with Sprootal?

Any of them, over MCP or the REST API. Claude, Cursor, n8n, an orchestrator you wrote yourself. Sprootal supplies the LinkedIn actions and the inbox. The intelligence stays yours.

Connect an agent. The Sandbox is free forever.

Read the docs

Talk to a founder

Thirty minutes with one of the two people who build this.