Codex + LinkedIn

Connect LinkedIn to Codex

Point OpenAI Codex at Crispy's LinkedIn MCP server. Search, message, and automate LinkedIn from the Codex CLI - agentic LinkedIn, no extension.

Get started Setup docs

Setup in 2 Minutes

1
Connect LinkedIn and copy your API key
Sign up, connect your account from the dashboard, and copy your key (it starts with crispy_)
2
Add Crispy to config.toml
Edit ~/.codex/config.toml and paste the block below with your key
3
Restart Codex
Start a new session. Codex auto-discovers every LinkedIn tool - ask it to do anything that touches LinkedIn.

~/.codex/config.toml

[mcp_servers.crispy]
command = "npx"
args = [
  "-y",
  "mcp-remote",
  "https://crispy.sh/api/mcp",
  "--header",
  "Authorization: Bearer YOUR_API_KEY"
]

What You Can Do with Codex + LinkedIn

Describe the LinkedIn task in plain language. Codex calls the right tools - or writes the script that does.

Outreach from the CLI

Run prospect search, draft personalized notes, and send connection requests from the Codex CLI. Describe the target and it calls the right tools.

“Find CTOs at Series A startups and send each a personalized connection request”

Data pulls and enrichment

Ask Codex to pull LinkedIn profiles, company data, and conversation history, and enrich your lead lists straight from the terminal.

“Enrich this CSV of leads with their LinkedIn profile data and company size”

Automated LinkedIn workflows

Chain search, messaging, and follow-up into a single agent task. Codex runs the workflow; Crispy keeps it inside safe limits.

“Message everyone who accepted my request this week and flag replies that sound like buying signals”

Inbox triage

Let Codex surface the messages that need you. Pull unread threads, rank the warm ones, and draft replies in your voice.

“Pull my unread LinkedIn messages and flag the ones that look like warm leads”

Why MCP Instead of a REST API?

When you connect Crispy via MCP, Codex doesn't just get API access - it gets tool awareness. It knows what each LinkedIn tool does, what parameters it takes, and how to chain them into a workflow.

Ask it to “find prospects and send personalized outreach” and it will use search_people, get_profile, and send_invitation in the right order - without you writing a single API call.

You still get full REST access too. Crispy supports both MCP and standard HTTP requests through the same endpoint.

Pricing

Crispy is $49/seat/mo (or $39/seat/mo annual). One plan, everything included. Works with Codex out of the box.

View full pricing

Frequently Asked Questions

Does OpenAI Codex support MCP servers?

Yes. Codex configures MCP servers in ~/.codex/config.toml. You add Crispy under an [mcp_servers.crispy] table and its LinkedIn tools become available to the agent.

Why use the mcp-remote bridge for Codex?

Crispy authenticates with an Authorization: Bearer header. The mcp-remote bridge connects to the remote Crispy endpoint and injects that header for you, which is the dependable way to attach a header-authenticated remote MCP server from config.toml.

Where is the Codex MCP config?

In ~/.codex/config.toml. Add the [mcp_servers.crispy] table shown above with your Crispy API key, then start a new Codex session.

Do I need Node installed?

The mcp-remote bridge runs via npx, so you need Node.js available. Any client that can send a custom header can also point directly at https://crispy.sh/api/mcp.

What LinkedIn tools are available in Codex?

Every Crispy tool: search people and companies, get profiles, send messages and connection requests, triage your inbox, create and schedule posts, pull analytics, and manage campaigns.

The complete LinkedIn API. Ready when you are.

Connect your first LinkedIn profile in under 5 minutes. Every tool, every seat, no feature gates. Safe limits, warm-up, and full permission control built in.

Connect your first profile

Read the docs