# Answer LinkedIn posts about your open-source project with Codex Source: https://finchreach.com/blog/codex-linkedin-library-mentions Published 2026-10-28 by Finchreach team. Tags: codex, linkedin, mcp, open-source, developer-relations. Use Codex to find LinkedIn posts that mention your library, check the facts against your repo, and queue accurate comments for your approval. ## The short answer Codex can watch LinkedIn for posts about your project and help you answer them with facts from your own repository. It searches posts with `linkedin_search`, reads the code and docs in your working directory to check what is true, and drafts a comment with `linkedin_comment_on_post`. With approval mode on, which is the default, each comment waits in your [approval queue](https://finchreach.com/approvals) until you approve it. The connection comes from Finchreach, a remote MCP server for LinkedIn (MCP is a standard plug that lets AI apps use outside tools). Codex supports remote MCP servers with a bearer token, so setup is three lines of TOML. ## Why would a maintainer answer LinkedIn posts from Codex? Developers talk about tools on LinkedIn more than many maintainers expect: “we migrated to X”, “has anyone used Y with Postgres?”, “Z dropped support for Node 18?”. Some of those posts are wrong in ways a single comment from the maintainer would fix. Most go unanswered because nobody on the team is searching LinkedIn. Codex is a good fit because it has the source of truth open. When a post claims your library does not support something, Codex can look at the code, the changelog and the docs before drafting a reply, and link to the right page. - Correct outdated claims with a link to the current docs. - Answer “how do I…” questions with the real configuration option, not a guess. - Thank people who wrote about your project, briefly and specifically. - Spot recurring confusion that should become a docs fix or an issue. ## How does the workflow run? 1. Start Codex in your project’s repository so it can read the code and docs. 2. Ask it to search LinkedIn posts for your project’s name and common misspellings with `linkedin_search` and the `content` vertical. 3. It lists the posts it found: author, a short summary, and whether the post contains a question, a claim, praise or nothing actionable. 4. For each post worth answering, it checks the claim against the repo and drafts a comment under the 1,250-character limit, usually much shorter. 5. You review the drafts in the terminal, then it queues the ones you accept. Each returns `pending_approval`. 6. You approve them in Finchreach. Codex can confirm the result with `linkedin_get_action`. If the same misunderstanding comes up several times, ask Codex to open a docs change for it in the same session. The LinkedIn comment fixes one reader; the docs fix the rest. ## Which prompts work for finding and answering LinkedIn mentions? > Ask your assistant: “Search LinkedIn posts (content vertical) for "fastqueue" and "fast-queue". List the 15 most recent with author headline, a one-line summary and a category: QUESTION, WRONG CLAIM, PRAISE, OTHER. Do not comment yet.” > Ask your assistant: “For each WRONG CLAIM, check the claim against this repo (src/, docs/, CHANGELOG.md). Tell me whether it was ever true and since which version it changed. Then draft a friendly comment under 60 words with a link to the relevant docs page.” > Ask your assistant: “For the QUESTION posts, draft answers that include the exact config option or function name from the code. If you are not sure, say so and leave that post out.” > Ask your assistant: “Queue the comments I approved here with linkedin_comment_on_post and give me the action ids. Then list any docs gaps these posts revealed.” Ask Codex to show its evidence (file and line) for each correction. It keeps the comment honest and makes your review fast. ## How do I connect Codex to LinkedIn? 1. [Create a Finchreach account](https://finchreach.com/sign-up) and connect your LinkedIn account on the Accounts page. 2. Create an API key on the API keys page and export it in your shell: `export FINCHREACH_API_KEY=lk_...`. 3. Add the server to `~/.codex/config.toml`, or to a trusted project’s `.codex/config.toml`, as below. You can also run `codex mcp add finchreach --url https://mcp.finchreach.com/mcp` and then add the token line. 4. Start Codex and ask “get my LinkedIn profile” to confirm the connection. 5. See the [Codex guide](https://finchreach.com/guides/codex), the [Codex integration page](https://finchreach.com/integrations/codex) and the [docs](https://finchreach.com/docs) for details. ~/.codex/config.toml: ```text [mcp_servers.finchreach] url = "https://mcp.finchreach.com/mcp" bearer_token_env_var = "FINCHREACH_API_KEY" ``` `bearer_token_env_var` makes Codex send `Authorization: Bearer ` and takes precedence over OAuth; `codex mcp login` is only for OAuth servers, so you do not need it here. Source: [Codex MCP docs](https://developers.openai.com/codex/mcp), as of October 2026. > Do not put the key in a project config that is committed to git. Keep it in the environment and reference the variable name. ## How should a maintainer comment on LinkedIn without sounding like a bot? - Say who you are in a few words: “maintainer here”. - Lead with the answer, then the link. One link is enough. - Correct claims gently. The author may have been right about an older version. - Skip posts that do not need you. Not every mention deserves a comment. - Do not comment on competitors’ posts to promote your project. - Keep the volume human: a handful of useful comments, not a sweep of every mention. ## What are the risks? > Automating a personal LinkedIn account carries risk under LinkedIn’s User Agreement. Finchreach’s approval queue and a human pace lower that risk; they do not remove it. There is also a content risk. LinkedIn posts are untrusted input, and one could contain text written to steer an assistant. Codex only drafts; nothing reaches LinkedIn until you approve it, so read each comment as if you were typing it yourself. ## Frequently asked questions Can Codex react to posts instead of commenting? Yes, with `linkedin_react_to_post`. Reactions also go through approval. Can it delete a comment I regret? Yes, `linkedin_delete_comment` removes one of your comments, after approval. Can it comment as our company page? No, Finchreach works with personal accounts only. Can it run this every morning on its own? Finchreach has no scheduler, and every comment needs your approval anyway. Run it when you have ten minutes. More workflows for developers: [release posts from Claude Code](https://finchreach.com/blog/claude-code-linkedin-post-from-changelog) and [answering DMs from Cursor](https://finchreach.com/blog/cursor-linkedin-dm-replies-from-codebase).