Skip to main content
Templates are pre-built agents with their instructions, context files and helper scripts. Use one as a starting point and customize it for your team. There are two ways to use a template:

​
Fork a template during onboarding

New DataGen Cloud accounts see two options when they start: Deploy from GitHub and Start from a Template.
1

Connect GitHub

Forking needs the DataGen GitHub App installed on your account, because the template is forked into your GitHub. If it is not installed yet, connect GitHub first.
2

Choose a template

Pick Start from a Template, then browse Choose a Template. Filter by category — Prospecting, CRM Hygiene, Lead Routing, Monitoring or Development. Each card shows its difficulty and the MCP servers it needs.
3

Use it

Click Use Template. DataGen forks the template’s repository into your GitHub account, connects the fork, and continues onboarding with connecting tools.
The fork is yours: change the agent by pushing to it, and set its runtime (vault, tool bundle, model, knowledge and data) on the agent page like any other repository agent. See Connect a GitHub repository.
Template forking uses GitHub and is not available on self-hosted installations. There, upload the template’s folder with Upload on the Agents page instead. See Upload an agent.

​
Install a template with Claude Code

The DataGen plugin for Claude Code installs templates from the DataGen agent templates repository.

​
Browse templates

Run it without arguments to see the catalog: each template’s description, required tools and MCP connections.

​
Install a template

1

Download files

Downloads the template’s files:
  • .claude/agents/<template-id>.md — the agent definition
  • .datagen/<template-id>/ — context, scripts, memory and working files
2

Check prerequisites

Lists the MCP connections, secrets and Python packages the template needs, and what is missing.
3

Install dependencies

Installs the Python packages listed in the template’s manifest.
4

Post-install guide

Shows what the agent does, how to invoke it, and any remaining setup.
After installing, run claude -r to reload Claude Code and pick up the new agent.

​
After installing

  • Connect the MCP servers it needs. Run /datagen:add-mcps, or add them to a vault in the web app. See Tools.
  • Add the secrets it needs as credentials in a vault. See Vault.
  • Run it in Claude Code: @<template-id> run the full pipeline.

​
Customize it

Installed templates are fully editable:
  • Context files in .datagen/<template-id>/context/ — your domain knowledge
  • Scripts in .datagen/<template-id>/scripts/ — data processing logic
  • Agent definition in .claude/agents/<template-id>.md — reasoning and workflow
A locally installed template is a Claude Code sub-agent in your project. To run the same workflow as a DataGen agent, put its instructions in the agent’s CLAUDE.md — for example by asking the Agent Builder chat to adapt the template — and publish it. See Build an agent.

​
Reusable skills

Skills are smaller, reusable capabilities that any agent can use. Install one locally with:
Skills install to .claude/skills/<skill-name>/. In DataGen, skills are uploaded to the workspace registry, reviewed, and then installed on agents. See Skills.

​
Next

Build an agent

Create and configure an agent in the Agent Builder

Apps

Build notebooks and custom apps on the same data and context