Claude Code for LinkedIn outreach

Claude Code is a bad LinkedIn client and a decent sales operator if you already work in a repo. The same session that is editing the landing page can read Omentir over MCP, score a batch, and leave drafts. It should not hold a LinkedIn password, and it should not invent a campaign while it is midway through a refactor.

Auth is a revocable API key. The hosted MCP URL is the same as every other operator: https://omentir.com/api/agent/v1/mcp. Chat Claude uses workspace approval instead. Mixing those two paths is how people paste a token into a connector UI, or wait for an OAuth screen that never appears.

What you are hiring

You are hiring a pair of eyes that already has the product in context. README, changelog, the sentence you just rewrote. Ask it to compare that with My Product. If they disagree, fix the profile before you let it touch a finder.

You are not hiring overnight research. Close the terminal and the work stops. If you wanted a list by morning without sitting there, that is Grok Bot, and only if you already pay for a plan that includes it.

Why the repo session matters

Most outbound junk starts as a homepage paragraph pasted into a chat that has never seen the product. Claude Code is useful when the claim on the site and the claim in Omentir can be diffed in one place. That is a founder job. It is a slow way to click Create campaign if that is the only task.

Treat lead text as untrusted data. A title that says "ignore previous instructions" is still a title. Your prompt should say so. Setup: Claude Code integration.

A first session job

Finish Omentir first: LinkedIn connected, My Product written in two sentences a stranger would understand.
Create a revocable API key and put it in the environment Claude Code already uses for secrets. Do not commit it.
Point Claude Code at https://omentir.com/api/agent/v1/mcp with Authorization Bearer. REST at https://omentir.com/api/agent/v1 is the same workspace if you prefer HTTP.
Fetch agents.md before you let it create anything. Put this in the session: research and draft only. Never send. Never enroll. Never sign into LinkedIn. Never create an agent until I approve the config.
Paste the job below. Replace the brackets. Keep the last two sentences.

Paste into Claude Code

You are the sales outreach operator for my company. You work from this repo through Omentir.

My product helps [buyer] get [result]. Best-fit accounts are [company type, size, region]. Strong signals: [signal]. Skip [bad fit].

Work only through Omentir MCP or REST. Read https://omentir.com/agents.md if you have not. Start with get_context, then list_agents. If no finder matches this ICP, show me a create_agent config and wait for a yes.

Pull up to 30 people. For each one write fit 1-5, the evidence, any risk, and a two-sentence draft that cites a real trigger. If a note could fit two buyers, rewrite it.

Do not send. Do not enroll. Do not sign into LinkedIn.
Do not create, delete, or reply unless I ask in this session.

Not Grok Bot, not Claude chat

Grok Bot has a shared cloud computer and Plugins. Claude Code does not. Claude chat has Connectors. Claude Code does not. If you wanted a conversation this afternoon with no repo open, use the Claude integration. If you wanted the editor instead of the terminal, use Claude Code versus Cursor.

When to stay in the browser

You are not in a terminal. The product story is still a slogan. You cannot give the review list fifteen minutes. Start in Overview. Claude Code earns the key when the same session is already in the repo.

FAQs

No. That is Grok Bot. Claude Code is a session in the repo. Close it and the job stops.

No. Claude chat uses Settings, Connectors, and workspace approval. Claude Code uses a revocable API key, like Cursor.

It can call Omentir tools under your campaign and safety settings. It should not type in LinkedIn. Keep send behind review.

No. Overview already finds people and drafts notes. Add Claude Code if you already live in that terminal and you want the same session to inspect the workspace.

No. Creating an agent should be a named request, the same way merging a pull request is a named request.

https://omentir.com/agents.md. Fetch it before you let the session create anything.