Run Airtop Agents From Claude Code, Codex, and Any Coding Agent

Mar 12, 2026

•

Daniel Shteremberg

•

product

If you're using Claude Code, Codex, or another AI coding agent to build and ship software, you now have a direct line to your go-to-market automation stack.

Today we're releasing the Airtop Agents Skill, an open-source extension that connects your Airtop agents to Claude Code, Codex, Cursor, Windsurf, and 40+ other coding agents. Install it in one command, and you can trigger lead enrichment, prospecting, account research, and any other GTM workflow you've built in Airtop, directly from your terminal.

Why This Matters

Airtop agents are purpose-built for GTM workflows. They run reliably, execute the same way every time, and cost a fraction of what runtime LLM agents charge. They can automate any website (including LinkedIn, vendor portals, and sites behind logins) using a real browser, and they combine that with API integrations and native enrichment services.

Until now, you'd run these agents from the Airtop portal or via webhooks. With this skill, your coding agent becomes a powerful orchestration layer on top of Airtop. It can call agents, process their results, chain multiple agents together, and weave GTM data into whatever you're building, all in a single conversation.

The combination is natural: Airtop handles reliable, cost-effective execution of browser-powered GTM workflows. Your coding agent handles the reasoning, orchestration, and data processing around them.

How It Works

Getting started takes about 30 seconds:

Install the skill:

npx skills add https://github.com/airtop-ai/airtop-skill --skill airtop-agents

Or if you use Claude Code specifically:

claude skill add --from https://github.com/airtop-ai/airtop-skill

Your coding agent will guide you to set up your API key.

That's it. Your coding agent now has access to every Airtop agent in your account.

What You Can Do

Once installed, you can interact with your agents using simple commands:

List your agents: /airtop-agents list
Run an agent: /airtop-agents run "Lead Enrichment" --vars '{"linkedin_url": "..."}'
Check status: /airtop-agents status <agentId> <invocationId>
View history: /airtop-agents history <agentId>

But the real power isn't in the commands themselves. It's in how your coding agent uses them. You can describe what you want in natural language, and the coding agent will figure out which Airtop agents to call, what data to pass, and how to process the results.

In Action: LinkedIn Engagement Mining

To show what this looks like in practice, let's walk through a real GTM workflow we ran entirely from Claude Code, using an Airtop agent that runs reliably and costs pennies per execution.

The goal: Find potential customers by identifying people who engage with thought leaders and competitors on LinkedIn.

The technique: This is a well-known go-to-market strategy. You identify important voices on LinkedIn (thought leaders, competitors, analysts) who talk about topics relevant to your business. Then you find everyone who commented on their posts. These commenters are actively engaged in your space, which makes them high-quality prospects.

How we did it:

We used this Airtop template, a free, ready-to-use agent for exactly this workflow. From Claude Code, we ran it with a single command, passing in the LinkedIn profile URLs we wanted to mine.

The agent found recent posts from the specified profiles, extracted every commenter, and enriched each one with:

Name and role
Company
Email (when available)
Link to the specific post they commented on

Claude Code received all of this structured data and displayed it as a clean table: names, companies, roles, emails, and post links, all ready for outreach.

The agent itself is highly configurable. You can have it push results directly to Google Sheets, send outreach emails, or even send LinkedIn DMs, all without leaving Airtop. Pair that with Claude Code orchestrating the flow, and you have a full prospecting pipeline: identify, enrich, and reach out.

Why this combination is powerful: The Airtop agent runs the same way every time, reliably extracting commenter data and enrichment at a fraction of what a runtime LLM agent would cost. Claude Code handles the reasoning around it: deciding what to do with the results, chaining additional steps, or processing the data further.

This is one agent and one use case. Browse the Airtop template library for dozens more, or build your own in minutes. You can chain multiple agents together, pass data between them, and build full GTM pipelines, all orchestrated by your coding agent, all running at predictable cost.

What You Can Build

This skill unlocks a new pattern: using your coding agent as the orchestration layer for your GTM automation stack. Here are a few ideas:

Lead generation pipelines. Run enrichment agents on a list of prospects, score and filter the results, then draft personalized outreach. Each Airtop agent execution costs pennies and runs deterministically, so you can process hundreds of leads without worrying about cost or inconsistent results.

Signal-based outbound. Identify prospects based on real engagement signals (like LinkedIn comments or job postings), enrich them with Airtop's native integrations, and have your coding agent craft tailored messaging for each one.

Account research. Before a sales call, ask your coding agent to run your Airtop research agent on the prospect's company. The agent handles the data gathering reliably across websites and APIs, while your coding agent synthesizes the results into a one-page brief.

Multi-step GTM workflows. Chain multiple Airtop agents together in sequence: one to identify prospects, another to enrich them, a third to score them against your ICP. Your coding agent ties them together and handles the logic between steps.

Airtop Agents as Composable Building Blocks

This release reflects how we think about Airtop agents: they're reliable, cost-effective GTM building blocks that can be called from anywhere. The Airtop portal is great for building and configuring agents, but you shouldn't be limited to running them only from there.

With this skill, your agents become tools that any coding agent can pick up and use. Build the agent once in Airtop (using natural language, no code required) and then invoke it from Claude Code, Codex, Cursor, or any other agent that supports skills. Each execution runs deterministically in Airtop's cloud, so you get consistent results whether it's the first run or the thousandth.

The economics make this practical at scale. Airtop compiles your natural language instructions into optimized workflows that minimize AI usage at runtime, so agent runs cost 10-100x less than runtime LLM agents. You can run prospecting, enrichment, and research agents continuously without worrying about unpredictable token costs.

Get Started

Install the skill in your coding agent of choice
Get your API key from portal.airtop.ai/api-keys
Browse the template library to find agents for your use case, or build your own in minutes using natural language

The skill is open source and available on GitHub: airtop-ai/airtop-skill

We'd love to hear what you build with it. Drop us a note or share your workflows. We're always looking for new use cases to feature.