Ollama · Model providerThrough another MCP clientWhat can Ollama do on LinkedIn with Finchreach?
Ollama doesn’t connect to MCP servers itself. Use it inside an MCP client such as Goose, Claude Code or Codex, add Finchreach there with your API key, and the same models can read your LinkedIn feed, messages and notifications, search LinkedIn, and draft posts, comments, messages and connection invitations, with every write waiting for your approval.
What can Ollama do on your LinkedIn?
Ollama provides the model, not the app that calls tools. Put it inside an MCP client and the same model can read your LinkedIn and draft for you, with every write waiting for your approval.
Ask your assistant“With a local model, summarize my LinkedIn messages without sending them to a cloud model.”
Ask your assistant“Draft a LinkedIn post from my notes using a local model.”
Ask your assistant“List my new LinkedIn connections from this week and draft a welcome message for each.”
| Reads (run right away) | Writes (wait for your approval) |
|---|---|
linkedin_list_accounts: The LinkedIn accounts connected to your workspace | linkedin_create_post: A text post |
linkedin_get_profile: A LinkedIn profile | linkedin_send_dm: A direct message |
linkedin_get_feed: Your home feed | linkedin_send_invitation: A connection invitation, with an optional note |
linkedin_get_notifications: Your notifications | linkedin_react_to_post: A reaction on a post (like by default) |
linkedin_get_conversations: Your message threads | linkedin_unreact: Removes your reaction from a post |
linkedin_get_messages: The messages in a thread | linkedin_comment_on_post: A comment on a post |
linkedin_search: Search everything, posts, people or companies | linkedin_delete_comment: Deletes one of your comments |
linkedin_get_received_invitations: Coming soon: the connection invitations waiting for your answer | linkedin_accept_invitation: Coming soon: accepts a received invitation |
linkedin_get_action: The status of a write: waiting for approval, approved, rejected or done |
What happens when Ollama drafts something?
- Ollama calls a write tool, for example
linkedin_create_postwith the text it wrote. - Finchreach doesn’t post it. It stores a pending action and tells the assistant it is waiting for approval.
- You open Approvals, read the exact text and the account it would go out from, and approve or reject it.
- Approved actions run on LinkedIn; rejected ones never do. Ollama can check the outcome with
linkedin_get_action.
Approval mode is on by default for every workspace. It is what makes it reasonable to let an assistant draft in your name.
Does Ollama work with Finchreach today?
| Ollama | |
|---|---|
| Remote MCP servers | None |
| Finchreach (OAuth or API key) | Through another MCP client |
| How sure we are | Medium: official docs incomplete, partly third-party sources |
| Checked on | 2026-10-28 |
Depends on the client: the client sends the Finchreach key, Ollama only runs the model. The field names for ollmcp and Open WebUI come from their projects, not from Ollama.
Finchreach’s server, https://mcp.finchreach.com/mcp, uses Streamable HTTP (pasting the bare https://mcp.finchreach.com works too). It signs in with OAuth 2.1, where you log in to Finchreach and approve a LinkedIn account and scopes, or with an API key sent as Authorization: Bearer lk_… or X-API-Key.
Which client should I run Ollama in?
- Goose: connects, with conditions. Setup guide.
- Claude Code: connects directly. Setup guide.
- Codex: connects directly. Setup guide.
The step-by-step Ollama guide explains why, with the exact screens and what to do instead.
Common questions
Not by itself, as of 2026-10-28. Depends on the client: the client sends the Finchreach key, Ollama only runs the model. The field names for ollmcp and Open WebUI come from their projects, not from Ollama.
No. With approval mode on, the default, every write Ollama makes through Finchreach (posts, comments, reactions, messages, invitations) becomes a pending action. You approve or reject it under Approvals before anything happens on LinkedIn.
Not through Finchreach today: posts are text only, there is no scheduling, and company pages and analytics aren’t supported. Writes happen when you approve them.
Any automation of a personal LinkedIn account carries some risk under LinkedIn’s User Agreement, and no tool can promise otherwise. Finchreach lowers it by keeping a person in the loop: nothing goes out until you approve it, and actions happen at your pace, not in bulk.
Third-party facts checked on against Ollama docs, github.com.