Skip to main content
Use this page when your agent can use a shell. You will install Deepline, authorize one workspace, and ask for one concrete GTM result.
Use Claude plugins if you prefer a Claude plugin. Use MCP setup if your agent does not use a shell.

Watch the setup

1. Install

Prerequisites

  • Claude Code or any AI coding agent
  • Permission for the agent to run setup commands

Ask your agent to install Deepline

Paste this into Claude Code, Codex, Cursor, Gemini CLI, or another shell-capable coding agent:
The agent handles Node.js and npm compatibility when the environment allows it, installs the CLI and Deepline skills, starts browser authorization, and verifies the result. If the runtime cannot be changed safely, the agent gives you the exact official install handoff and resumes afterward. In a restricted sandbox it automatically uses a persistent folder-local install.You only need to approve the agent’s native command prompt and the Deepline browser authorization.
If this command reports that Node.js or npm is missing or unsupported, install the current Node.js LTS release from nodejs.org and retry, or use the coding agent flow above to handle runtime setup.If setup fails, use the exact retry or diagnostic command in its JSON response. Run deepline doctor --json when setup directs you to diagnose the environment.

2. Run one request

The Deepline skill auto-syncs when you install the CLI. In Claude Code, type:
The /deepline-gtm prefix tells your agent to use Deepline. Describe the result and constraints after the prefix. Start with one of these requests:
Or describe your own business problem. Your agent will turn it into one concrete /deepline-gtm prompt, show the reusable prompt, and run it.
A useful first run ends with one of two outcomes: a result with the fields you asked for, or a clear reason no result was found. Review that outcome before you run a larger list.
The waterfall tries providers in order. It stops after a valid result. Billing depends on the current pricing unit for each provider. Some providers charge for a result. Other providers charge for a request or page. Examine the route before you scale. Run a representative pilot. Coverage varies by ideal customer profile and region.

What happened

  1. Your agent selected a prebuilt Play or created a custom Play.
  2. Deepline ran the selected work and recorded the run.
  3. The result shows the requested fields, or why a result was unavailable.
  4. You now have a concrete prompt to reuse or refine.

3. Scale only after the first result looks right

For a list, use this request:
Deepline reads the file and selects the applicable waterfall. Then, Deepline enriches each row. Pilot mode: Ask the agent to test representative rows before the full file. Include easy, typical, and difficult inputs. These differences can change coverage or cost. Examine the provider route and measured pilot cost. Then, approve or reject the full run. For the checklist, read Cost control.

Next steps

Deepline Skills

Choose the packaged GTM skill for research, enrichment, review, or reusable automation.

API Reference

How tools, payloads, and waterfall execution work under the hood.

Database Access

Query resolved customer data from the local database surface.

Email Waterfall Recipe

Step-by-step: find verified emails for a list of contacts.

Provider Integrations

Per-provider guides, schemas, and sample outputs.

I Have X, I Want Y

Pick the right workflow based on what data you already have.

Example Skills

Open-source GTM workflow templates on GitHub.

Reach us: /deepline-feedback in Claude Code, or team@deepline.com.