Now in private beta · join the waitlist

Outbound on autopilot
that learns from every reply.

Describe the outcome in one sentence. Agentic Outreach finds the right people, researches them, writes every touch across email, LinkedIn and InMail, handles replies, books meetings, and rewrites its own playbook from what prospects actually answer.

From one sentence to booked meetings. No sequence builder, no spreadsheets.

The status quo

Outbound today is a junkyard of tools and a tired founder.

A lead database, an enrichment tool, a sequencer, a LinkedIn extension, a shared inbox, a calendar link and a spreadsheet to glue them together. Every tool does one step. Nobody learns anything, least of all the tools.

  • Spreadsheet SwampLists exported, cleaned and re-imported by hand.
  • Template Factory“Hi {first_name}” at scale, and prospects can tell.
  • Channel SilosEmail and LinkedIn don’t know about each other.
  • Inbox PitReplies triaged at 11pm, meetings lost in threads.
  • Guesswork Dept.Campaign 12 is exactly as smart as campaign 1.
Isometric illustration of a tired founder on a stool surrounded by stacked spreadsheets, tangled monitors, a broken conveyor spilling envelopes and a leaking pipe

How it works

One agent. The whole outbound motion.
In a closed loop.

Five stations, connected. What happens at the end feeds the start, so every campaign begins smarter than the last one ended.

Isometric miniature city: a research lab with a satellite dish, a writing studio with a robotic arm, a signal tower beaming in three directions, a mail sorting station, and an observatory with a rising chart, all connected by pipes that loop back to the start

Inside each station

Built like a senior SDR team.
Priced like software.

The Lab research

Finds people worth writing to, and says why.

Sources companies against your ICP, picks the two decision-makers who matter, and reads funding news, hiring, tech stack and their recent posts. Every prospect gets a fit score with its reasons shown, not hidden.

94 fit · Mara Kellerhiring 3 growth roles · posted on embedded payments 6d ago · Series B

The Studio write

Writes like it read their posts. Because it did.

Every message is drafted from that prospect’s own signals, in your voice, signed as you. Edit any of them before launch.

The Tower send

Three channels, one conversation.

LinkedIn-first, email-first or InMail, chosen per person. A LinkedIn accept stops the cold email; a reply anywhere stops everything.

LinkedInEmailInMail

The Sorting Room reply

Replies handled. Meetings on your calendar.

Every reply is classified: interested, question, out of office, not now, bounce. The agent drafts the answer in context and books the slot.

  • InterestedThursday works → invite sent
  • QuestionPricing + case study drafted
  • Out of officePaused until the 14th

The Observatory learn

A playbook that rewrites itself.

Openers, channels, timing and audience slices are tested continuously. Winners get promoted across every route; losers retire quietly.

Multi-channel

LinkedIn and Gmail.
One conversation.

Most tools run email and LinkedIn as two campaigns that never talk to each other. Here every prospect gets one sequence across both, and each channel knows what just happened on the other.

Mara Keller · VP Growth, PaywiseRoute picked because: posts weekly, 2nd-degree connection
  1. Day 1Connection requestSent without a note
  2. Day 2AcceptedOpener goes out 5 min later, citing her post on embedded payments
  3. Day 2Cold email cancelledShe’s already talking on LinkedIn
  4. Day 4Reply: “Thursday?”The agent confirms a slot in your voice
  5. Day 4Meeting bookedThu 15:00 CET invite from your Google Calendar

Illustrative sequences

From your own LinkedIn profile.

  • Connection requests, direct messages and InMail
  • An opener goes out minutes after someone accepts
  • The agent can carry a thread through to a booked meeting
  • Daily invites and messages stay inside LinkedIn’s limits

Connect once. Your profile and posts also teach the agent your writing voice.

From your real Gmail inboxes.

  • Connect Gmail or Google Workspace with Google sign-in
  • Sends rotate across warmed mailboxes, 20 a day each
  • Replies, bounces and out-of-office notices read from the thread
  • Hard bounces stop the prospect for good, never retried

Zoho Mail and any SMTP inbox work too.

The learning loop

Week 6 knows things week 1 didn’t.

Every send is an experiment with a labeled outcome. The agent keeps a running scoreboard per persona, channel and message pattern, then shifts traffic to what works. Your team never has to run an A/B test again.

Hypothesize
“Referencing a recent post beats referencing company news for growth leaders.”
Split
Traffic divided across live variants, inside your daily sender limits.
Measure
Positive replies, accepts and meetings. Opens don’t count.
Promote
The winner becomes the default everywhere. A new challenger enters.
Experiment boardWeek 1

Opener

Company news0.0%
Recent post reference0.0%

collecting replies…

Positive reply rate2.1%
Wk 1Wk 2Wk 3Wk 4Wk 5Wk 6

Illustrative model of the loop

MCP server

Run outbound
from Claude.

Agentic Outreach ships with its own MCP server. Paste one URL into Claude, ChatGPT, Codex or Cursor, approve the connection, and your assistant can set up your account, plan campaigns, research prospects, review warm leads and check results in plain language, without opening the dashboard.

  • Claude
  • ChatGPT
  • Claude Code
  • Codex
  • Cursor
  • Any MCP client
No send tool, by design
Your assistant can research, plan, draft and approve. Messages only go out through the campaign you launched.
Scoped, revocable access
Sign in with OAuth, no key to paste. Separate read, draft, campaign and approve scopes, and one click disconnects any agent.
Spends only when you say
Audience research runs free against the existing prospect database. Paid sourcing waits for your OK.
agentic-outreachconnected

Any warm leads this week? Draft openers for the best two.

find_warm_leadsmin_score: 70→ 3 leads
  • 94Mara Kellerposted about embedded payments
  • 88Tom Weissraised a Series A on Monday
  • 71Ana Ruizhiring 2 SDRs
draft_outreach×2→ preview only · nothing sent

Both drafts are ready. Mara’s opens with her post from Tuesday. Approve them into the Paywise campaign?

Approve Mara. Hold Tom for now.

approve_warm_leadMara Keller→ enrolled · sends with the campaign

50 tools, grouped by job

streamable HTTP · OAuth 2.1 or API key · a selection shown

Understand

  • whoami
  • list_connected_accounts
  • list_industries
  • preview_audience

Plan & build

  • plan_campaign
  • create_campaign
  • run_discovery_gap
  • discovery_status

Research & write

  • list_campaign_prospects
  • research_prospects
  • get_dossier
  • draft_message

Warm leads

  • find_warm_leads
  • get_lead_evidence
  • draft_outreach
  • approve_warm_lead

Watch & steer

  • campaign_insights
  • list_campaigns
  • get_campaign
  • list_agents
  • pause_agent

Guardrails

Autonomous, not unsupervised.

You approve the launch

The agent researches and drafts on its own. Launching, paid expansion and risky changes wait for a person.

Sender-safe by design

Warmed mailboxes rotate at 20 sends a day each. LinkedIn invites stay inside platform limits.

Stops the moment it should

A reply, bounce or opt-out halts the sequence on every channel. Hard bounces are never retried.

Every decision explained

Fit scores, channel choices and exclusions show their inputs. No black box, no “trust us”.

Pricing

Priced per prospect worked.
From $99 a month.

One credit takes a prospect from sourcing to the end of their sequence. No seats to count, no sequences to meter, and your own AI assistant is included on every plan.

See full pricing

14-day trial with 100 credits · extra LinkedIn senders $49/month · private beta, by waitlist

Describe the outcome.
The agent does the rest.