n8n Releases Desk, an Open Source AI Agent Interface With Built-In Security and Audit Controls.

n8n has released Desk, an open source interface for running AI agents that prioritizes security and transparency over convenience. Unlike tools like Claude Desktop that execute agent actions directly on a user's machine, Desk routes every operation through a user's own n8n instance, keeping credentials encrypted and every step visible in an audit trail.

The project emerged from a real-world failure. Marcel Claus-Ahrens, who introduced Desk in a recent episode with Dylan Watkins, described having to apologize to a client after Claude Code performed an unauthorized action. That incident drove the design philosophy behind Desk: give AI agents the flexibility to act, but never let them operate outside observable, controlled boundaries.

Desk looks and behaves like Claude Desktop, presenting users with a familiar chat interface. The difference is architectural. When an AI agent in Desk needs to perform an action, like reading a file, calling an API, or querying a database, that request is routed through the user's n8n instance rather than executed locally.

This means credentials never leave the n8n environment. Every action becomes a workflow step, complete with inputs, outputs, and timestamps. If something goes wrong, teams can trace exactly what the agent did and when. For organizations running AI agents in production, that auditability matters.

Desk is model-agnostic. It works with any language model, including local deployments. Users can swap between OpenAI, Anthropic, or self-hosted models without changing the interface. The project is free and open source, with the code available on GitHub.

Alongside Desk, n8n introduced a Model Context Protocol (MCP) server that lets AI agents build workflows for the first time. Previous integrations allowed agents to trigger existing workflows. This update gives them the ability to construct new ones from natural language prompts.

As previously covered, the MCP server works across AI clients that support the protocol, not just Desk. Users can describe a task in plain language, and the agent will generate a multi-step workflow, connect the necessary nodes, and configure parameters. The workflow remains editable in the standard n8n interface, allowing technical teams to refine what the agent produces.

This capability positions n8n differently than pure no-code platforms. Where tools like Zapier or Make focus on pre-built templates and point-and-click configuration, n8n now offers a hybrid: conversational workflow creation with full code-level access when needed.

AI agents are moving from demos to production. When that happens, teams need more than speed. They need accountability, security, and the ability to debug when things fail. Desk and the MCP server address those needs without forcing users to abandon the flexibility that makes AI agents useful in the first place.

For developers building on n8n, this release follows the company's recent momentum, including their $5.2 billion valuation and SAP investment. The platform is positioning itself as infrastructure for teams that need both the power of code and the accessibility of visual tools, now extended to AI agents that can build their own automations.

Desk is available now as an open source project. The MCP server ships as part of n8n's standard distribution.