> ## Documentation Index > Fetch the complete documentation index at: https://support.affinity.co/llms.txt > Use this file to discover all available pages before exploring further. # Warm Intro Agent — overview > Surface ranked paths to a target and draft the ask on request — run on demand from AI Chat or the affinity-warm-intro MCP skill. Getting a warm intro is something you already do by hand: figure out who at your firm knows the person, then ask them to connect you. The Warm Intro Agent does the legwork — it surfaces the strongest paths to your target from your firm's relationship data and drafts the ask when you request it. You run it on demand from **AI Chat**, and it's also available as the `affinity-warm-intro` **MCP skill** in your own LLM client (Claude, ChatGPT, etc.). It **never sends anything on your behalf.** Brand Warm Intro ## What it does 1. **Surfaces ranked paths.** Given a target person or company, it identifies who at your firm has the strongest relationships and returns candidate paths ranked by strength. 2. **Drafts the ask on request.** When you ask, it prepares an intro request for the connector to forward, ready for you to review. 3. **Hands it back to you.** You take the draft and make the intro yourself — the agent doesn't send. ## A typical flow ### Step 1 — Name your target In AI Chat (or via the `affinity-warm-intro` MCP skill), tell the agent who you want to reach: > “I'd like an intro to the founder of Greenfield Systems.” ### Step 2 — Review the ranked paths The agent surfaces the strongest routes through your firm's network, ranked, with the evidence: ```text theme={null} Best path to Greenfield Systems (founder) Jordan Lee — strongest relationship Why: emailed last month · shared board seat Alternate: Priya Shah (met at conference, weaker) [ Use Jordan ] [ Use alternate ] ``` Pick the connector that makes sense. If neither fits, choose someone else. ### Step 3 — Ask for a draft When you're ready, ask the agent to draft the ask for your chosen connector to forward: ```text theme={null} To: Jordan Lee “Would you be open to introducing me to [founder] at Greenfield Systems? We're exploring their space and your connection would mean a lot. Happy to send a forwardable blurb.” ``` Edit the wording to sound like you. **The agent doesn't send — you send it yourself** through your normal channel. ### Step 4 — Send it yourself Take the draft and send the ask to your connector the way you normally would. Making the intro stays in your hands. 1. You want an intro to a founder at Greenfield Systems, so you ask the agent in AI Chat. 2. The agent surfaces the best internal paths, ranked — e.g. *“Jordan Lee has the strongest relationship (recent email + shared board).”* 3. You ask it to draft the ask for Jordan to forward. 4. You review and edit the message. 5. You send it yourself through your normal channel. ## How to stay in control * **You run it on demand.** The agent doesn't monitor in the background — it acts when you ask, in AI Chat or through the MCP skill. * **It drafts; it doesn't send.** No automatic outreach. * Each suggested path shows *why* it's a strong route (the relationship evidence). * It respects permissions — it only uses connections your account can already see. ## Common questions * **Will it email anyone automatically?** No. It surfaces paths and drafts on request; you send the message yourself. * **How do I run it?** Ask a warm-intro question in AI Chat, or use the `affinity-warm-intro` skill via the Affinity MCP. * **How does it rank the paths?** Relationship strength across your firm's interaction data — and it shows that evidence. * **What if the suggested path is wrong?** Choose another suggested connector, or edit the draft before you send. ## Coming later Intro-request tracking, follow-up drafting, and background monitoring are future direction — the launch experience surfaces paths and drafts the ask on request. For the full detail on what powers your paths (explicit relationships, LinkedIn, and inferred connections), see [Warm Intros in AI Chat and MCP](/s/article/Warm-Intros-in-AI-Chat-and-MCP). *** *Affinity AI can make mistakes. Review the path and the draft before you send.* ## Related topics - [Warm Intros in AI Chat and MCP](/s/article/Warm-Intros-in-AI-Chat-and-MCP.md) - [Affinity Ascend agents — Overview](/s/article/Affinity-Ascend-agents-overview.md) - [Working with Affinity agents](/s/article/Working-with-Affinity-agents.md) - [AI & MCP](/s/article/AI-and-MCP.md) - [Set up & configure your agents](/s/article/Set-up-and-configure-your-agents.md) - [How to prioritize warm introductions and travel planning](/s/article/How-to-prioritize-warm-introductions-and-travel-planning.md) - [Prompts for Affinity AI Chat](/s/article/Prompts-for-Affinity-AI-Chat.md) - [Enabling Affinity Ascend — An Admin's Guide](/s/article/Enabling-Affinity-Ascend.md) - [Find warm introductions to a target](/s/article/Find-warm-introductions-to-a-target.md)