Gemlead in your CRM, your stack and your AI.

An X lead generation API lets your own tools receive the buyers found on X, and act on them, without anyone exporting a list by hand. Gemlead's API gives you the leads its agents find and score on X, each with its score and the reason for it, plus the personas, campaigns, conversations and DMs behind them. Signed webhooks call your server the moment a lead turns hot or replies, and an MCP server lets Claude, ChatGPT or Cursor run your Gemlead in plain words. All three come with the Business plan, on one key.

The point is speed: a hot lead that waits in a tab until someone copies it into the CRM is a lead going cold. Below: what that delay costs, what each piece does, and a real request.

Why your X leads need to reach your stack

Leads that stay in a separate tool get worked late, or not at all.

• Disconnected tools slow everything. Over half of sales leaders with AI, 51%, say disconnected systems are slowing down their AI initiatives, and 74% of sales professionals are focusing on data cleansing, according to Salesforce's State of Sales 2026, a survey of 4 050 sales professionals.
• Copying data eats selling time. The average seller spends 40% of their time selling. Gen Z reps are at 35%, losing about two full hours a week to manual data entry, in the same report.
• Hot leads cool fast. In a Harvard Business Review study of 1.25 million leads, companies that tried to reach a lead within an hour were nearly seven times as likely to qualify it as those that waited longer.
• Research is moving into AI. Gartner predicts that by 2027, 95% of sellers' research workflows will begin with AI, up from less than 20% in 2024.

What Gemlead's X lead generation API, MCP server and webhooks do

The REST API

Every endpoint lives under https://gemlead.ai/api/v1 and speaks JSON. In Gemlead, open API in the menu, pick the X account, name the key and choose its scopes, read alone or read and write. The key starts with gl_live_, shows once, and only ever sees that X account's data.

What it covers:

• Leads: list them with filters like hot, score_min, a persona or a date, read where a lead came from, rescore it, follow it, or save your own metadata on it, like a CRM ID.
• Personas and sources: create a persona, add sources, change settings, share it with your other X accounts.
• Campaigns: create a draft from a template, start or pause it, read its stats and trends day by day.
• Conversations: read the threads, send a DM with text, an image, a GIF, a video or a voice note, react, move a conversation in your pipeline.
• Exports and analytics: a CSV or JSON Lines export of your leads, and the charts of your Dashboard.
• Julien AI: ask Gemlead's AI assistant about your account from your code.

Your hot leads, newest first:

curl -G https://gemlead.ai/api/v1/leads \
  -H "Authorization: Bearer $GEMLEAD_API_KEY" \
  -d hot=true \
  -d limit=10

Each lead comes back with its X profile, its score from 0 to 100, its score_reason in one sentence, its persona and its status. A key makes 120 requests a minute. DMs sent through the API are capped at 10 a minute and 300 a day per X account, or less when X sets a lower quota.

Webhooks

Register a public HTTPS URL in the app under API, or with POST /v1/webhook_endpoints, and pick your events:

Each request carries a Gemlead-Signature header, an HMAC SHA256 you check with your endpoint's secret. A failed delivery is retried 7 times, over about 45 hours. You can keep up to 10 endpoints per account, and limit one to the leads of some personas. The webhooks guide shows the check in Node.js and Python.

The MCP server

The Model Context Protocol is an open-source standard for connecting AI applications to external systems. Gemlead's server at https://gemlead.ai/mcp gives Claude, ChatGPT, Cursor and Claude Code over 60 tools, one for each thing the API does: search_leads, get_campaign_stats, list_conversations, send_message, create_export and more. Claude and ChatGPT sign in with OAuth, with no key to copy. In Claude Code, one command:

claude mcp add --transport http gemlead https://gemlead.ai/mcp \
  --header "Authorization: Bearer $GEMLEAD_API_KEY"

Then ask: "Show me today's hot leads for my Tipsters persona, with the reason each one scored high." Write tools need the write scope, and the assistant asks you before it sends, starts or deletes anything. To work on several X accounts, add one connection per account. The full setup is in the docs.

Why it works

• Push beats polling. A webhook reaches your server the moment a lead turns hot, so your CRM or your rep gets it inside that first hour, without spending requests to check.
• One record per lead. Save your CRM ID in the lead's metadata, and every later event links back to the same record. Fewer duplicates, less cleanup.
• AI assistants already speak MCP. Anthropic reports more than 10 000 active public MCP servers and over 97 million monthly SDK downloads, with MCP adopted by ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code. Gartner expects 60% of B2B seller work to run through conversational AI interfaces by 2028.
• Agents need good data. Gartner predicts that by 2028, fewer than 40% of sellers will say AI agents improved their productivity. Feeding yours scored leads with a reason for each score, rather than a raw list, is how you end up among the sellers who see the gain.

Manual exports vs the Gemlead API

Who the API is for

• Developers and RevOps who sync X leads into HubSpot, Salesforce, Pipedrive, Attio, Notion or Google Sheets, with their own code or an automation tool that receives webhooks.
• Founders who work in Claude or ChatGPT and want their hot leads and replies in the same chat.
• Agencies that report to clients from the analytics endpoints, with one key per client X account.
• Teams building their own AI agents, who give them Gemlead's scored leads and its DM tools instead of a scraper.
• B2C sellers, like a sports tipster or a creator with a paid community, who want each hot lead in their team's own tool.

How those leads are found is in how to find leads on X. Before an agent writes to them, read what AI SDRs can and cannot do.

Best practices

1. One key per tool, kept on your server in an environment variable like GEMLEAD_API_KEY. Give write only to the tools that change data.
2. Verify every webhook. Check the signature on the raw body and refuse timestamps more than 5 minutes old.
3. Answer fast, work later. Reply 2xx within 10 seconds, then process the event in a queue. Delivery is at least once, so skip an event id you already handled.
4. Export for bulk. One export beats hundreds of pages.
5. Read before you allow. An assistant can be tricked by text it reads. Check what a write tool is about to do before you approve it.

FAQ