# Read-only LinkedIn market research with Nous models and Hermes Agent Source: https://finchreach.com/blog/nous-linkedin-market-research Published 2026-10-28 by Finchreach team. Tags: nous, hermes, linkedin, mcp, research, use-case. Use a Nous model through Hermes Agent and Finchreach to research what a market is posting about on LinkedIn, without sending a single write. ## The short answer You can have a Nous model read LinkedIn for you: search posts and companies on a topic, summarize what people in a market keep talking about, pull out recurring pain points and list which companies are most active. Done this way, nothing is ever posted, sent or liked. It is pure reading, which makes it the lowest-risk way to start using an AI assistant on LinkedIn. One honest caveat first. As of October 2026, the Nous Portal and its chat have no documented MCP client, so they can’t connect to Finchreach directly. The realistic route is Hermes Agent, Nous Research’s open-source agent: sign it into the Portal with `hermes setup --portal` so it uses Nous models, and add Finchreach as an MCP server in `~/.hermes/config.yaml`. MCP (Model Context Protocol) is a standard plug that lets AI apps use outside tools. Finchreach is a remote MCP server at `https://mcp.finchreach.com/mcp` that operates your own LinkedIn account, authenticated with an API key you create in the Finchreach dashboard. ## Why start with read-only LinkedIn research? Most of the worry about AI on LinkedIn is about writing: a post you didn’t mean to publish, a message with the wrong tone, an invitation sent to the wrong person. Research avoids all of that. The assistant only calls read tools, so there is nothing to approve and nothing that goes out under your name. It is also genuinely useful. LinkedIn is where many B2B buyers, founders and practitioners talk about their work in public. Reading a few hundred posts by hand to find the patterns takes an afternoon; an agent can do the first pass in minutes and hand you a summary with links you can check. Even if you later decide to let the assistant draft posts or messages, a few weeks of read-only use is a good way to learn how it reasons and where it gets things wrong, before anything it writes can reach real people. ## Which Finchreach tools does market research use? Only the read side of the server: - `linkedin_search` with the `content` vertical finds posts about a topic. Results page with a cursor, so the agent can keep reading past the first batch. - `linkedin_search` with the `companies` vertical finds companies by keyword, with page-based paging. The `people` vertical finds individuals, and `all` mixes result types. - `linkedin_get_profile` reads a profile, which helps the agent understand who is behind a post (a founder, a buyer, a vendor). - `linkedin_get_feed` reads your own feed, a useful sanity check on what your network is seeing. You never need the write tools for this job. Simply don’t ask for posts, comments, messages, reactions or invitations. And if the agent did try one by mistake, approval mode is on by default for every Finchreach account: the write would come back as `pending_approval` and sit in your queue at [/approvals](https://finchreach.com/approvals) until a person approved or rejected it. ## What do you ask the agent? Start broad, then narrow. A first pass on a market might look like this: > Ask your assistant: “Use only Finchreach read tools; do not create posts, comments, messages or reactions. Search LinkedIn posts about "accounts payable automation" and read at least 50 results. Summarize the five themes people post about most, with two example post links for each theme.” Then dig into pain points, which are usually the most valuable output: > Ask your assistant: “From those posts, list the recurring complaints or problems people describe in their own words. Group them, quote short phrases where possible, and note whether each one comes from buyers, vendors or consultants based on the authors’ profiles.” And map who is active: > Ask your assistant: “Search LinkedIn companies for "AP automation" and "invoice processing software". Cross-check against the authors of the posts you read. Give me a table of the companies that appear most often, what they seem to emphasize, and one post link per company.” Finish with something you can act on: > Ask your assistant: “Write a one-page brief: top themes, top pain points, most active companies, and three gaps nobody seems to be talking about. Mark anything you are unsure of.” ## How do you connect a Nous model to LinkedIn? 1. Create a Finchreach account at [finchreach.com/sign-up](https://finchreach.com/sign-up), connect your LinkedIn account under Accounts, and create an API key under API keys. 2. Install Hermes Agent and run `hermes setup --portal` to sign it into the Nous Portal, so it uses Nous models. 3. Add Finchreach under `mcp_servers` in `~/.hermes/config.yaml` (snippet below), then run `/reload-mcp` or restart Hermes. 4. Ask the agent to list your LinkedIn accounts. If it returns yours, the tools are live. 5. See the [Nous guide](https://finchreach.com/guides/nous), the [Hermes guide](https://finchreach.com/guides/hermes) and the [quickstart](https://finchreach.com/docs) for details. ~/.hermes/config.yaml: ```text mcp_servers: finchreach: url: "https://mcp.finchreach.com/mcp" headers: Authorization: "Bearer " ``` > Use `url` or `command` for a server in Hermes, not both. Keep the API key out of shared dotfiles and repos; anyone with it can use your LinkedIn account through Finchreach, subject to your approval queue. ## How do you keep the research trustworthy? - Ask for links. A summary you can’t check is an opinion; one with example post links is evidence. - Ask the agent to say how many posts it actually read. Search results are a sample, not the whole market. - Watch for vendor noise. In many markets, a lot of posts are written by companies selling to the buyers you care about. Ask the agent to separate the two. - Repeat the same prompts every few weeks and compare. A change in themes over time is often more telling than any single snapshot. - Treat what you read as public posts by real people. Use the findings for your own thinking, not to copy someone’s wording. ## Is reading LinkedIn with an agent risk-free? No. It is lower risk than writing, but automating any activity on a personal LinkedIn account carries risk under LinkedIn’s User Agreement. Finchreach acts as your own account, at a pace you set by how you use it. Keep research sessions reasonable in size rather than asking the agent to read thousands of results in one go. > The approval queue covers writes. For reads there is nothing to approve, so the habit that matters is asking for sensible amounts of data. ## Questions people ask Can I connect Finchreach to the Nous Portal chat directly? Not as far as we can tell. As of October 2026 there is no documented MCP client in the Portal or its chat. Use Hermes Agent with the Portal as its model provider. Does this post, comment or message anything on LinkedIn? No. The prompts above only use search, profile and feed reads. Any write the agent attempted would wait for your approval. Can it read analytics for companies or posts? No. Finchreach doesn’t provide LinkedIn analytics or company page access. It reads what your own account can see: search results, profiles, your feed, notifications and messages. Where do I go next? Follow the [Hermes guide](https://finchreach.com/guides/hermes) for setup, or read [what a LinkedIn MCP server is](https://finchreach.com/blog/what-is-a-linkedin-mcp-server) if you are new to the idea.