# How to track LinkedIn outreach via API
> Use a REST API or MCP client to monitor pending invitations, reply intent, and send windows so you can adjust LinkedIn outreach pacing before account health slips.
- HTML: https://getkinlead.com/scaling-outreach-with-external-api-tools
- Markdown: https://getkinlead.com/scaling-outreach-with-external-api-tools.md

**Short answer:** To track LinkedIn outreach via API, connect a script or MCP client to a provider that manages the LinkedIn account and exposes invitation, reply, and send-window data. Pull pending invitations, acceptance rate, and reply intent on a regular schedule, then slow down or pause when the pending pool grows or replies turn negative. Reading data is lower risk than sending through an API, so keep sends inside the limits and send window you set.

## What to track before you touch an API

Total messages sent tells you almost nothing about account health. A quiet account can be closer to a restriction than a busy one if most invitations sit unanswered. The numbers that matter show whether people are accepting and replying. Whether your automation is running is a separate question.

Pending invitations are the first metric. When requests pile up, LinkedIn has less evidence that your outreach is wanted. You don't need a published limit to act on this. Watch the trend in your own account, and when the pending pool grows for several days, stop new invites until it clears.

Reply intent is the second metric. A reply that says 'not interested' is still a reply, but it's a different signal from 'send me pricing.' Read intent in batches, not one message at a time. If negative or neutral replies start crowding out interested ones, your targeting or message is off, even if the send volume looks normal.

| | Metric | What it tells you | Warning sign |
| --- | --- | --- | --- |
| Pending invitations | Requests that have not been accepted | The pool grows for several days in a row |
| Acceptance rate | How often people accept your request | A drop that lasts longer than a normal slow week |
| Reply intent | Whether replies are interested, neutral, or negative | Negative replies start showing up more often |
| Send window | When your actions actually go out | Messages land outside the lead's working hours |

## How to connect a script or MCP client to LinkedIn data

LinkedIn does not offer a public API for personal outreach. You need a provider that connects to your LinkedIn account and exposes the data through a REST API or an MCP server. That provider should handle the LinkedIn connection without storing your password in your script.

Once you have a provider, authenticate with OAuth for chat apps or an API key for custom scripts. Keep the key in a password manager or an environment variable. The goal is to move pending invitations, acceptance rate, reply intent, and send-window data into one place you check on a schedule.

1. **Select a provider.** Look for a tool that connects to your LinkedIn account and exposes a REST API or MCP server. Ask whether it stores your LinkedIn password. The answer should be no.
2. **Authenticate your account.** Use OAuth for chat apps or an API key for scripts. Keep the key out of shared documents and screenshots.
3. **Test a read endpoint.** Pull recent conversations and pending invitations before you build anything. Confirm the data matches what you see in LinkedIn.
4. **Map the data.** Send the fields you need to a sheet or dashboard: pending count, acceptance rate, reply intent, and send window. Don't pull everything just because the endpoint returns it.

## Reading reply intent without opening every thread

An API gives you text, not judgment. The useful part is grouping replies by intent so you can see whether the account is attracting buyers or just attention. Start with three buckets: interested, neutral, and negative. Add a fourth for 'asked to check back later' if that keeps showing up.

Run the classification on a small sample first. If the AI puts a polite 'no' in the interested bucket, fix the prompt before you trust the summary. You are looking for a trend, not a perfect label on every message.

Keep the output short enough to act on. A weekly summary with counts by intent and a few example replies is more useful than a spreadsheet with every message tagged.

> Read replies in batches before you change pacing. One angry reply is noise. A week of negative replies is a signal.

```
Ask your MCP-connected client: 'Read my recent LinkedIn replies. Group them into interested, neutral, negative, and check back later. Show two example replies for each group. Do not draft responses yet.'
```

## Pacing changes that keep the pending pool healthy

Pacing is a response to data. Treat it that way. If your pending pool is growing, you are sending faster than your audience is accepting. Stop new invites, withdraw the oldest ones if your provider supports it, and wait for the pool to come down.

When acceptance is steady and the pending pool is low, you have room to add volume. Do it in small steps. If you double your sends because one week looked good, you won't know which change caused the next problem.

Use the send window as part of pacing. A message that goes out at 2 a.m. in the lead's time zone can look automated even when a person wrote it. Set a window that matches how your buyers work, then keep it stable long enough to compare results.

**Guessing pacing**

- Sending a fixed number of invites every day regardless of replies
- Increasing volume because you have a long lead list
- Ignoring pending invitations until the account is restricted

**Data-driven pacing**

- Lowering volume when pending invitations grow for several days
- Increasing volume only after a week of steady acceptance
- Pausing new invites when reply intent turns negative

## A weekly API monitoring routine

A daily check is too much for most small teams, and a monthly check is too late. A weekly routine catches the trends that matter without turning account health into a full-time job.

Keep the same four checks each week. The order matters less than doing them consistently, because you are comparing this week with last week, not judging a single snapshot.

1. **Monday: check the pending pool.** Compare the current pending count with last week. If it grew, cut new invites until it returns to your normal range.
2. **Wednesday: read reply intent.** Pull the week's replies and group them. Look for negative replies that mention tone, timing, or a bad fit.
3. **Friday: review send windows.** Check whether any messages went out outside the window you set. Fix the window before you change volume.
4. **Monthly: export and compare.** Export leads and replies to CSV. Compare the last four weeks. Look for changes in acceptance and intent, not just total sends.

## Doing this with Kinlead

Kinlead has a REST Agent API and an MCP server, so Claude, ChatGPT, Cursor, Claude Code, Codex and other MCP clients can find leads, check replies and draft answers. Chat apps connect with OAuth; scripts use API keys. You can read more about [the agent API and MCP server](/features/agent-api-and-mcp).

Kinlead also watches account health inside the app. Settings shows pending invitations and acceptance rate with warnings. Stale pending invitations are withdrawn automatically to keep the pending-invite pool healthy. When LinkedIn pushes back on an account, Kinlead pauses invites on that account and emails you.

Pacing stays under your control. Each LinkedIn account sends at most one action every 5 minutes, inside daily invite and message limits you set and inside a send window in the lead's time zone. Replies from every campaign land in one inbox, classified by intent, with interested replies first.

> Use hand-off mode when you want the conversation to come to you at the first reply.

## Related

- [Agent API and MCP server](https://getkinlead.com/features/agent-api-and-mcp.md)
- [LinkedIn account safety](https://getkinlead.com/features/linkedin-account-safety.md)
- [Book a demo](https://getkinlead.com/demo.md)

## Common questions

**Does using an API increase the risk of account restriction?**

Reading data through a provider is lower risk than sending messages through a raw API. The account restriction risk comes from how fast you send, how many invitations sit unanswered, and whether the provider keeps outreach inside limits you set.

**How often should I pull health data?**

Check pending invitations and reply intent once a day or once a week, depending on your volume. The important part is comparing the trend, not checking every hour.

**Can an MCP client draft replies from the health data?**

Yes. Give the client your sales playbook, the reply text, and the intent label. Ask for a draft, then decide whether to send it yourself or let the campaign handle replies.

**What should I do when pending invitations grow?**

Stop new invites, withdraw stale requests if your provider supports it, and wait for the pool to come down. Don't raise volume to make up for a slow week.

[Create a Kinlead account](https://getkinlead.com/signup) or [book a call](https://getkinlead.com/demo) to talk through pricing and setup.
