# A daily LinkedIn digest with any OpenRouter model Source: https://finchreach.com/blog/openrouter-linkedin-notification-digest Published 2026-10-28 by Finchreach team. Tags: openrouter, linkedin, mcp, digest, notifications. OpenRouter isn’t an MCP client, so pair it with the OpenRouter Agent SDK or an MCP-capable client and Finchreach to get a daily LinkedIn digest. ## The short answer OpenRouter gives you one API for many models, but it doesn’t connect to MCP servers on its own. To use an OpenRouter model with Finchreach, run it through something that speaks MCP: OpenRouter’s Agent SDK with the `@openrouter/mcp` helper, or any MCP client that accepts OpenRouter as its model provider (such as Goose, Cline, OpenCode or LangChain). A practical first project is a daily LinkedIn digest: a short script reads your notifications and messages, and a model of your choice summarizes what needs you today. Reading changes nothing on LinkedIn. Any replies it drafts go to your Finchreach [approval queue](https://finchreach.com/approvals) and only send after you approve them. MCP (Model Context Protocol) is a standard plug that lets AI apps use outside tools. Finchreach is a remote MCP server for LinkedIn at `https://mcp.finchreach.com/mcp`, authenticated with an API key header. > OpenRouter also runs its own MCP server at `mcp.openrouter.ai`. That is a separate server from OpenRouter and unrelated to this setup: it isn’t what connects your models to Finchreach. ## Why build a LinkedIn digest instead of checking the app? LinkedIn’s notification tab mixes things that need you (a reply from a lead, a comment asking you a question, a connection request from someone you met) with things that don’t (likes, work anniversaries, suggested posts). Opening the app to find the first kind usually means losing ten minutes to the second. A digest flips that. Once a day, you get a short message: who wrote to you, which comments asked a question, which invitations look relevant. You open LinkedIn only when something in the digest is worth it, or you approve a drafted reply without opening it at all. OpenRouter fits this job because you can choose the model per task and change it without rewriting anything: a cheaper fast model for summarizing, a stronger one for drafting replies. ## Which Finchreach tools does the digest use? - `linkedin_get_notifications` for notification cards and the unseen counts. - `linkedin_get_conversations` for the inbox (participants and last message), and `linkedin_get_messages` to read a thread in full. - Optionally `linkedin_send_dm` or `linkedin_comment_on_post` to draft replies, which go to your approval queue. ## How do I connect an OpenRouter model to Finchreach? The full walkthrough is in the [OpenRouter guide](https://finchreach.com/guides/openrouter). With the Agent SDK: 1. [Create a Finchreach account](https://finchreach.com/sign-up), connect your LinkedIn account on the Accounts page, and create an API key (it starts with `lk_`). 2. Get an OpenRouter API key and install the OpenRouter Agent SDK and `@openrouter/mcp`. 3. Call `createMCPTools()` with the Finchreach URL, the `streamableHttp` transport and your key as a bearer token. 4. Pass `mcp.tools` to `callModel` with the model you want, then close the connection when done. digest.ts: ```ts import { createMCPTools } from '@openrouter/mcp'; const mcp = await createMCPTools({ url: 'https://mcp.finchreach.com/mcp', transport: 'streamableHttp', auth: { kind: 'bearer', token: process.env.FINCHREACH_API_KEY! }, }); // callModel({ model: '', input: DIGEST_PROMPT, tools: mcp.tools }); await mcp.close(); ``` The exact `callModel` setup depends on your Agent SDK version; follow [OpenRouter’s MCP tools docs](https://openrouter.ai/docs/agent-sdk/call-model/mcp-tools), as of October 2026. Run the script on a schedule with cron or any job runner, and send the output wherever you read things: email, Slack or a file. Prefer not to write code? Use an MCP client that accepts OpenRouter as a provider (base URL `https://openrouter.ai/api/v1`), such as Goose, Cline or OpenCode, and add Finchreach as a remote MCP server with an `Authorization: Bearer` header. See the [Goose guide](https://finchreach.com/guides/goose) or our [quickstart](https://finchreach.com/docs). ## What prompt makes a useful LinkedIn digest? Ask for a fixed structure so the digest stays scannable: > Ask your assistant: “Use the Finchreach tools to read my LinkedIn notifications and my conversations with a new message from the other person. Write a digest with three sections: Needs a reply (name, what they asked, one line), Worth a look (comments or mentions that ask me something), and New connections (people who accepted my invitations, name and headline). Skip likes, anniversaries and anything that needs no action. Under 250 words. Don’t send or queue anything.” If you want drafts as well, run a second step with a stronger model: > Ask your assistant: “For each item under Needs a reply, draft a short reply in a friendly, direct tone and queue it with linkedin_send_dm in that conversation. Never promise prices or dates. List the action ids at the end.” Keeping the read-only digest and the drafting step separate makes it easy to run the digest every day and the drafts only when you have time to review them. ## Can the script send messages without my approval? Not with approval mode on, which is the default. Every write, whether a DM, comment, reaction, post or invitation, becomes a pending action with status `pending_approval`. You review the exact text on the [approvals page](https://finchreach.com/approvals) and approve or reject it. The script can check outcomes later with `linkedin_get_action`. When you switch models on OpenRouter, behavior changes too. One model may follow “don’t queue anything” perfectly; another may not. The queue means a model change can’t send anything you haven’t seen. > Automating a personal LinkedIn account carries risk under LinkedIn’s User Agreement. The approval queue lowers that risk because you review and pace every action, but it doesn’t remove it. A daily read-only digest is a light, human-paced pattern; keep it that way. ## OpenRouter and LinkedIn: common questions Can I add Finchreach to OpenRouter’s chat as a tool? Not as a hosted remote MCP tool. OpenRouter doesn’t call remote MCP servers for you; the MCP connection runs in your code or your client. Which models work? Any model on OpenRouter that supports tool calling. Test a few runs, since tool use varies a lot between models. Does my OpenRouter key give access to LinkedIn? No. The OpenRouter key pays for models; the Finchreach key, which you can revoke on the API keys page, is what reaches your LinkedIn tools. Want everything on your own machine instead? See [drafting LinkedIn posts with a local Ollama model](https://finchreach.com/blog/ollama-linkedin-local-drafts). New to MCP? Read [what a LinkedIn MCP server is](https://finchreach.com/blog/what-is-a-linkedin-mcp-server). Get your first digest tomorrow morning: [sign up for Finchreach](https://finchreach.com/sign-up) and follow the [OpenRouter guide](https://finchreach.com/guides/openrouter).