# Sixteen capabilities. Six specialists. One account.

URL: https://zevari.ai/linkedin-mcp/tools

> Every capability your AI sales team has, which specialist owns it, and whether it runs on its own or on your AI's call. Over MCP or REST. Every send waits for your approval until you switch a trained specialist to Autopilot.

What the team can do

Every one of them is listed here, with who owns it and whether it runs on its own.

Reach them over MCP from Claude, Codex, or any MCP client, or over the REST API from your own code. Research, scoring, and drafting can run without you. Sending never does.

[Build my AI team](https://zevari.ai/auth/signup?redirect=/app) [Connect over MCP](https://zevari.ai/linkedin-mcp)

Runs on its own a Zevari worker triggers it On your AI's call your AI asks for it, the result is saved to the lead

## Research and enrich 5

Everything the team learns about a person or a company before anyone writes a word.

1. **Company intelligence**
  Runs on its own
  Runs when a lead is enrolled. Cached 90 days.
  Prospecting`agents_company_intelligence`
2. **Person research**
  On your AI's call
  Profile, posts, comments, education, skills.
  Prospecting`agents_research_person`
3. **Behavioral profile**
  On your AI's call
  How they engage and what they respond to.
  Prospecting`agents_behavioral_profile`
4. **Communication blueprint**
  On your AI's call
  How to write to this person specifically.
  Voice Writer`agents_communication_blueprint`
5. **Lead enrichment**
  On your AI's call
  Verified email for the lead, or a phone number when explicitly asked.
  Prospecting`agents_enrich_lead`

## Score and classify 5

Sorting the list by what actually closed, and reading what comes back.

1. **ICP scoring**
  Runs on its own
  Scores new leads against your criteria, with reasons. Learns from your reviews.
  ICP Scoring`agents_icp_score`
2. **Signal scoring**
  Runs on its own
  Ranks the reasons to reach out as the scanner finds them.
  Signals`agents_score_signals`
3. **Signal classification**
  Runs on its own
  Sorts signals by kind so the right specialist picks them up.
  Signals`agents_classify_signals`
4. **Inbox radar**
  On your AI's call
  Reads a thread for intent and fit.
  Reply Qualifier`agents_inbox_radar`
5. **Search from plain English**
  On your AI's call
  Turns a sentence into a LinkedIn search.
  Prospecting`agents_extract_search_params`

## Write 4

Every draft in your voice, and every one of them staged for your approval.

1. **Message generation**
  Runs on its own
  Drafts each step when a lead is enrolled. Also on your AI's call.
  Voice Writer`agents_generate_messages`
2. **Connection notes**
  On your AI's call
  The note that goes with the request.
  Voice Writer`agents_generate_connection_note`
3. **Warming comments**
  On your AI's call
  A real comment on their post before the ask.
  Voice Writer`agents_generate_warming_comment`
4. **Personalization pass**
  On your AI's call
  Rewrites a draft against the research.
  Voice Writer`agents_personalize`

Two ways in

## Your AI, or your code.

Same sixteen capabilities, same account, same approval gate. Pick the door that matches how you already work.

### Over MCP

Add the hosted endpoint to Claude Code, Codex, ChatGPT, or any MCP client and the team shows up as tools your AI can call.

[Claude](https://zevari.ai/linkedin-mcp/claude) [Codex](https://zevari.ai/linkedin-mcp/codex) [ChatGPT](https://zevari.ai/linkedin-mcp/chatgpt)

### Over REST

Call the same capabilities from your own services when you would rather not go through a chat client at all.

[Quickstart](https://zevari.ai/quickstart)

Questions

## The simple version.

[The safety layer under all of it](https://zevari.ai/safety)

### What does "runs on its own" mean?

A worker Zevari owns triggers it, on a schedule or on an event like a new lead being enrolled. You do not have to ask. Everything else runs when your AI calls it, from Claude, Codex, or your own code, and the result is saved to the lead.

### Which specialist owns which capability?

Each row names its owner. Prospecting and Signals fill and rank the list, ICP Scoring sorts it, Voice Writer writes, Follow-up keeps it moving, Reply Qualifier reads what comes back. The capabilities are the tools they use.

### How does my AI get at these?

Over MCP for Claude Code, Codex, and any MCP client, or over the REST API from your own services. Same capabilities either way, same account, same approval gate on every write until a trained specialist moves to Autopilot.

### Does any of this send on its own?

No. Research, scoring, and drafting can run without you. Sending cannot. Every message, connection request, comment, and post waits for your approval, unless you have moved that specialist to autopilot after it has earned it.

### Will this get my LinkedIn account banned?

The safety layer sits under every capability here: a session-based connection with no browser cookies, enforced weekly ceilings and daily caps, working hours and burst limits, and the approval gate. See the safety page for the numbers.