# Follow up on LinkedIn after an event with Goose

Source: https://finchreach.com/blog/goose-linkedin-event-follow-up
Published 2026-10-28 by Finchreach team. Tags: goose, linkedin, mcp, events, networking.

Give Goose your conference notes and it finds the people you met on LinkedIn and drafts personal connection invitations, each queued for your approval.

## The short answer

Yes, Goose can handle your LinkedIn follow-up after a conference or meetup. Paste in your notes (names, companies, the talks you saw, what you talked about), and Goose looks up each person on LinkedIn, matches them to the right profile, and drafts a connection invitation with a short note that mentions your actual conversation. Nothing is sent until you approve it.

The connection runs through Finchreach, a remote LinkedIn MCP server at `https://mcp.finchreach.com/mcp`. MCP (Model Context Protocol) is a standard plug that lets AI apps use outside tools. Goose supports remote MCP servers natively as "extensions", so you add Finchreach once with your API key and the LinkedIn tools are there in every session.

Approval mode is on by default for every Finchreach account. Each invitation Goose drafts comes back with the status `pending_approval` and waits in your [approval queue](https://finchreach.com/approvals) until you approve or reject it.

## Why do event follow-ups on LinkedIn usually slip?

You come home with a phone full of half-names, a few business cards and a notes file that says things like "Priya, platform team, talked about flaky CI." The window where people remember you is short, and turning those scraps into twenty personal invitations takes an evening nobody has. So you send a few generic requests, or none.

The slow parts are finding the right profile (there are many people called Priya on platform teams) and writing a note that sounds like you remember the conversation. Both are good jobs for an assistant, as long as you keep the final say.

## What can Goose do with my LinkedIn through Finchreach?

For this workflow, Goose uses a small set of Finchreach tools:

- `linkedin_search` with the people vertical, filtered by network distance (1st, 2nd or 3rd+), to find candidates by name, company and role.
- `linkedin_get_profile` to check a candidate against your notes (current company, title, location) before drafting anything.
- `linkedin_send_invitation` with an optional note of up to 300 characters. The draft is queued for your approval, not sent.
- `linkedin_send_dm` for people you are already connected to, either in an existing conversation or as a new message.
- `linkedin_get_action` to check whether a queued action is still pending, approved or rejected.

Search can also cover content and companies, which helps when your notes only say "the speaker from the observability startup." Finchreach does not post images, schedule posts, manage company pages or provide analytics, so this workflow sticks to search, profiles, invitations and messages.

## What should I tell Goose after a conference?

Start with your raw notes. You do not need to clean them up; the more context about the conversation, the better the note.

> Ask your assistant: “I just got back from KubeCon. Here are my notes: Priya (platform team at a fintech, talked about flaky CI and test sharding), Marco Ruiz (gave the talk on eBPF tracing, I asked about overhead), Dana from the Seattle meetup who runs a DevRel team. For each person, search LinkedIn people, show me the top matches with title and company, and wait for me to confirm before drafting anything.”

Once you confirm the matches, ask for the invitations:

> Ask your assistant: “Draft a connection invitation for each confirmed person with a note under 300 characters. Mention the specific thing we talked about, keep it casual, no sales pitch, and sign with my first name. Queue them and list the action ids.”

Goose will hand back drafts along these lines, each one queued:

Example invitation note:
```text
Hi Priya, great talking test sharding with you at KubeCon. I tried the split-by-timing idea on our CI this week and it already cut the flaky retries. Would love to stay in touch. - Sam
```

For people you are already connected to, a short message works better than an invitation:

> Ask your assistant: “Dana and I are already 1st-degree connections. Draft a short DM thanking her for the DevRel chat at the Seattle meetup and asking if she would share the onboarding doc she mentioned. Queue it for approval.”

Later, check on the queue:

> Ask your assistant: “Check the status of the invitations you queued earlier and tell me which are still pending approval.”

## How do I connect Goose to LinkedIn?

1. Create a Finchreach account at [finchreach.com/sign-up](https://finchreach.com/sign-up) and connect your LinkedIn account in the dashboard under Accounts.
2. Create an API key under API keys. It starts with `lk_`. Treat it like a password.
3. In Goose Desktop, open Extensions, choose Add custom extension, pick the Streamable HTTP type, enter `https://mcp.finchreach.com/mcp` as the URL, and add a request header `Authorization` with the value `Bearer lk_...`.
4. In the Goose CLI, run `goose configure`, choose Add Extension, then Remote Extension (Streamable HTTP), and enter the same URL and header.
5. Start a session and ask Goose to list your LinkedIn accounts. If it returns your account, you are connected.

If you prefer to edit `~/.config/goose/config.yaml` directly, an entry like the one below is reported to work. Note that the `headers` key is not confirmed in the official Goose docs (bearer header support comes from a Goose pull request and third-party guides), so the Desktop or `goose configure` route is the safer one.

~/.config/goose/config.yaml:
```text
extensions:
  finchreach:
    name: finchreach
    type: streamable_http
    uri: https://mcp.finchreach.com/mcp
    headers:
      Authorization: Bearer lk_your_key_here
    enabled: true
    timeout: 300
```

> The `--with-streamable-http-extension` command-line flag does not take headers, so it cannot pass your API key. To send your key, use one of the routes above.

The full walkthrough is in the [Goose guide](https://finchreach.com/guides/goose), and the [docs](https://finchreach.com/docs) cover API keys, the approval queue and every tool. Goose’s own extension docs are at [block.github.io/goose](https://block.github.io/goose/docs/getting-started/using-extensions).

## Can I run this with a local model in Goose?

Goose has a documented Ollama integration, so you can run it on a local model if you prefer to keep your notes on your machine. The LinkedIn calls still go to Finchreach and LinkedIn, of course. The catch is tool use: this workflow chains several tool calls (search, profile check, invitation), and smaller local models often struggle to pick the right tool or fill in its arguments. Choose a model that is known to handle tool calling well, and expect to confirm matches more often.

## Is it safe to let an AI agent send LinkedIn invitations?

Every write Goose makes through Finchreach (invitations, messages, posts, comments, reactions) lands in your approval queue first. Open [/approvals](https://finchreach.com/approvals), read each note, fix the ones that sound off, and approve or reject them. The pace is yours: nothing goes out in bulk because the agent decided it should.

Matching is the step to watch. Common names, people who changed jobs and conference badges with nicknames all produce wrong matches. Asking Goose to confirm matches with you before drafting, as in the prompts above, catches most of them.

> Automating a personal LinkedIn account carries risk under LinkedIn’s User Agreement. The approval queue and a human pace lower that risk; they do not remove it. Keep invitations personal and the volume modest.

## Goose and LinkedIn event follow-up: common questions

How long can the invitation note be? Up to 300 characters. Ask Goose to stay under that and to mention one concrete detail from your conversation.

Can Goose send the invitations without me? No. Approval mode is on by default, so each invitation returns `pending_approval` and only goes out after you approve it in the dashboard.

What if I only remember a first name and a company? Give Goose that plus anything else you remember (role, talk title, city). It can search people and companies and show you candidates to pick from.

Can it follow up with people I am already connected to? Yes. It can draft a direct message in an existing conversation or start a new one, also queued for approval.

Does it work with other agents? Yes, any MCP client that can send an API key header. See the [Claude Code integration](https://finchreach.com/integrations/claude-code) or the [docs](https://finchreach.com/docs).