See SocialKaptan live — 15 min walkthroughBook demo

LinkedIn Networking Agent

The full warm outreach loop — engage, connect, engage again, wait, DM, one follow-up — without living in your inbox or sounding like a template.

Key takeaways

  • Find people via keyword search or paste profile links — same engage → connect → engage flow for both.
  • Pre-connect: comment on one post for context, send a short no-pitch connection note, then engage 1–2 more posts.
  • Post-accept: wait ~2 days (no pitch on accept), then one opening DM — specific observation + open question.
  • Exactly one soft follow-up after 5 quiet days, then the person is retired — no spam loops.
  • Five run modes: full daily loop, discover & connect, observe & comment, send due messages, or check replies.
  • Editable daily budgets: 8–10 warm-up comments, 20 connection requests, 10 intro DMs. Pending invites expire after 3 days.
  • Optional Book demo mode turns the DM thread over to an agentic closer that reads the conversation and decides what to send next.
  • Plan-based concurrent people limits (5 on Starter, 15 on Pro, 40 on Agency) keep volume conservative.

Cold connection requests and generic DMs rarely convert — but the warm version (find the right person, engage with their content first, connect with context, wait, then reach out with a real observation) takes more time than most teams have. The LinkedIn Networking Agent automates that full sequence at a human pace, in your own browser session.

The full workflow

  1. 1

    Find your person

    Keyword people search fills your pipeline automatically, or paste any linkedin.com/in/… profile link to queue them manually. Both paths use the same outreach sequence.

  2. 2

    Engage → connect → engage

    For each person: comment on one recent post (context for the note), send a short connection request with a no-pitch AI note referencing what you saw, then engage with 1–2 more posts if available. Pending invites are tracked until accepted or expired.

  3. 3

    Wait ~2 days after accept

    When they accept, the agent waits about two days before any DM — no pitch on accept. Engagement already happened during outreach, so this is a natural pause, not more spam.

  4. 4

    Opening DM (not a pitch)

    One personalized message: lead with them, one specific observation from their posts, an open question (never yes/no), no call ask. The AI is briefed on warm outreach principles — curiosity over conversion.

  5. 5

    One follow-up, then done

    If there's no reply within a few days, exactly one soft follow-up goes out — light, pressure-free, with an easy out. Then the person moves to completed whether they replied or not.

Outreach principles baked in

  • First message is about them, not you — connection notes reference their actual content.
  • Every message ends with an open question or low-pressure offer — never a yes/no opener or early call ask.
  • Bio and price come late — only when asked. Value and curiosity come first.
  • Exactly one follow-up if ghosted. Then move on — no guilt-trip loops.
  • Human tone: contractions, no template phrases, no "I hope this finds you well."

Five ways to run it

The full daily loop does everything in sequence, but each stage can also be run on its own — useful while you're tuning tone, or when you want to push one part of the pipeline forward without touching the rest.

Run modes
ModeWhat it does
Full daily loopDiscovery, warm-up engagement, connection requests, due DMs, and reply checks — worked across the day with safe spacing, then it waits for tomorrow
Discover & connect onceProcesses queued profile links and/or keyword search: engage a post, send the connection note, engage 1–2 more
Observe & comment onceChecks people inside the post-accept wait window and leaves warm-up comments
Send due messages onceSends the opening DM to everyone past the wait period, then stops
Check for replies onceLooks for responses on messaged people and advances, follows up, or retires them

Daily budgets and timing

Every number here is editable in Setup → Limits. The defaults are deliberately low — this is per-person outreach, and volume is the fastest way to make it feel like spam.

Default networking limits
SettingDefaultWhat it controls
Networking comments per day8–10Likes and comments on posts before connecting, plus wait-period engagement
Connection requests per day20LinkedIn invites with a note — continues even after the comment budget is spent
Intro DMs per day10Opening messages to people past their wait period
Days after accept before first DM2The deliberate pause between accepting and hearing from you
Days to wait before follow-up5Quiet days after the intro DM before the single soft follow-up goes out

There's a useful bit of behaviour in how those budgets interact: when today's comment budget is spent, discovery doesn't stop. It keeps observing and connecting without engaging posts first, up to the connect cap. You lose the pre-connect warm-up on those people, not the whole day's outreach.

The People and Visited tabs

Warm outreach only works if you can see where everyone stands. Four views cover that:

  • People → Active: everyone in the post-accept loop, with a status pill — Observing, Ready to message, Messaged, Followed up, Replied, or No reply.
  • People → Outreach queue: profiles waiting for their engage → connect run, plus a count of invites already sent and still pending.
  • Visited: every profile the agent has opened, so discovery never circles back to the same person twice.
  • History: a persistent record of runs and actions that survives restarts, alongside the live friendly log.

You can paste any linkedin.com/in/… link straight into the queue, and remove anyone from any list at any time. The active count is shown against your plan cap, and the input disables itself when you're at the ceiling rather than silently dropping people.

What the AI writes, and where you steer it

Three separate pieces of copy get generated per person, each with its own job:

  1. 1

    The warm-up comment

    A genuine reaction to something specific in their post. This is what makes the connection note make sense a few minutes later.

  2. 2

    The connection note

    Short, no pitch, referencing what you just saw on their profile or post. Its only job is to explain why you're reaching out.

  3. 3

    The opening DM

    One specific observation from their content plus an open question — never yes/no, never a call ask. Sent about two days after they accept, not the moment they do.

Setup → Style holds two guidance fields: comment guidance (how you sound when engaging) and message guidance (why you actually want to connect). Both sit on top of the shared writing style you configure once in Settings, so the agent's voice matches the rest of your output.

Book demo mode

Warm outreach eventually has to go somewhere. Book demo is an optional agentic closer: instead of a fixed follow-up script, an LLM reads the actual thread each time and decides what to send next, with booking a demo as the goal, using the Book demo URL from Settings.

It's off by default, and it's a per-run toggle as well as a saved default — so you can leave it off for pure networking and switch it on for a sales-focused batch. It works with the daily loop and with the message and reply modes.

Reliability on LinkedIn's fragile UI

The connect flow is the flakiest part of LinkedIn to automate: the Connect button hides behind a More menu on some profiles, "Add a note" moves, and the Send button changes state as you type. Selector-only automation silently misses these and sends noteless invites.

  • Visual matching: the agent compares the live screen against reference images of the Connect, More, Add-a-note, Send, and Pending states, and clicks by what's actually rendered rather than by a selector alone.
  • Post-send verification: after sending, it confirms the invite really went out and shows as pending, instead of assuming success.
  • Open-chat guard: LinkedIn's chat overlay and minimized message tabs intercept clicks, so the agent closes them before acting.
  • Stale invite cleanup: pending invites older than three days are dropped and the person moves on, so the queue doesn't silt up.

Separate from the Comment Agent

The Networking Agent runs in its own Chrome profile alongside the LinkedIn Comment Agent and Post Generator — so you can engage in public feed conversations and run per-person warm outreach at the same time, without either agent interfering with the other. Run All can start both, plus Instagram and Reddit, in one click.

Setup checklist

  1. 1Connect LinkedIn in the desktop app; the Networking Agent gets its own Chrome profile automatically.
  2. 2Fill in product details and (if you'll use Book demo) your Book demo URL in Settings.
  3. 3Write comment guidance and message guidance in Setup → Style — the second one matters most: say plainly why you want to connect.
  4. 4Add 3–6 keyword phrases in Setup → Targeting, or paste a handful of profile links into People.
  5. 5Run Discover & connect once and read what it posted and sent before enabling the daily loop.
  6. 6Review Setup → Limits, then start the full daily loop.

Frequently asked questions

The Comment Agent engages on public posts by keyword, feed, or URL — it's for visibility in conversations. The Networking Agent works a per-person pipeline: find someone, engage with their post, connect with a no-pitch note, engage 1–2 more posts, wait ~2 days after accept, then send an opening DM and at most one follow-up. They're complementary and run in separate browser profiles.

Concurrent people in the post-accept warm-up loop are capped by plan: 5 on Starter, 15 on Pro, and 40 on Agency. Queued profiles and pending connection invites don't count against the active cap until accepted.

Yes — paste any LinkedIn profile URL into the People tab to queue them. The agent runs the same engage → connect → engage sequence as keyword discovery, then moves them to active after they accept. Both sources feed the same pipeline.

Pending invites expire after three days and the agent drops them, so it never nags the same person or lets the queue fill with dead invites. If they accept later, they can be picked back up.

Twenty by default, editable in Setup → Limits. Warm-up comments (8–10/day) and intro DMs (10/day) have separate budgets, so running out of one doesn't halt the others.

Because pitching on accept is what makes LinkedIn DMs feel like a trap. The engagement already happened during outreach, so the gap reads as a natural pause rather than a sales sequence firing. It's configurable if you want it shorter or longer.

No. Five quiet days after the intro DM it sends exactly one soft follow-up — light, pressure-free, with an easy out — and then the person is retired to the completed list whether or not they replied.

An optional agentic closer. With it on, an LLM reads each DM thread and decides what to send next, aiming to book a demo using the Book demo URL from Settings, rather than following a fixed script. It's off by default and can be toggled per run or saved as your default.

Yes. Run Discover & connect once with a single profile queued, or use Observe & comment / Send due messages / Check for replies individually. Each mode processes one stage and stops, so you can read the actual output before committing to the daily loop.

It runs locally in your own Chrome profile with human-paced timing between actions, low daily budgets, a plan-capped number of concurrent people, and three-day expiry on stale invites. It also verifies each invite actually sent rather than firing blindly. No tool eliminates risk, but the volume here is closer to an attentive person than to a scraper.

The two agents track their budgets separately — networking warm-up comments come out of the networking comment range (8–10/day by default), not the Comment Agent's. Running both means both budgets are in play, which is worth accounting for if you're pushing limits upward.

See LinkedIn Networking Agent running on a real desktop

Local-first LinkedIn & Instagram automation with AI replies, preview mode, and safe daily limits.