---
{
  "title": "Build AI agents with the Workflowy API",
  "intro": "Use the Workflowy API directly when you need a custom agent, integration, or automation."
}
---

The Workflowy API gives your code direct access to Workflowy nodes. Use it when an existing MCP or CLI connector does not fit your product, hosting, authentication, or workflow requirements.

# What you can build

- An agent that captures research or meeting notes into a chosen Workflowy location
- A workflow that reads a project subtree and summarizes open work
- An integration that creates, updates, moves, completes, or searches nodes
- A hosted service that uses Workflowy as a structured workspace
- Custom tools for mirrors, backlinks, calendar nodes, imports, and exports

# Get started

{% steps %}
1. Get a key from [workflowy.com/api-key](https://workflowy.com/api-key).
2. Open the [Workflowy API reference](/api-reference/).
3. Follow a guide or choose the endpoint for the operation you need.
4. Authenticate requests with your API key.
5. Test reads and narrowly scoped writes before allowing an agent to make broader changes.
{% /steps %}

# Design agents for safe changes

AI agents can make mistakes. Give them the smallest useful scope, require confirmation for destructive actions, preserve stable node IDs, and verify important writes after they happen.

For an already packaged local connection, compare [Workflowy Desktop MCP](/help/mcp-server) and the [self-managed connector options](/help/run-your-own-connector).