GNOME UI MCP
Automate GNOME on Wayland—discover UI via AT-SPI and drive input through Mutter—from an MCP agent for desktop QA and workflow scripts.
Overview
GNOME UI MCP is an MCP server for the Build phase that automates the GNOME Wayland desktop using AT-SPI discovery and Mutter-driven input for agent tooling on Linux.
What is this MCP server?
- GNOME Wayland desktop automation combining AT-SPI element discovery with Mutter input injection
- Packaged as OCI image ghcr.io/asattelmaier/gnome-ui-mcp:0.4.1 with Docker stdio runtime hints
- Host uid/gid mapping and network host mode for faithful session integration on Linux
- Enables agent-driven UI walks on native GTK apps where browser MCP cannot reach
- Published server version 0.4.1 as OCI ghcr.io/asattelmaier/gnome-ui-mcp:0.4.1
- Docker runtime with host network and default uid/gid 1000 in schema
What problem does it solve?
Linux builders cannot let agents operate native GNOME applications because standard browser MCP leaves the whole desktop shell invisible.
Who is it for?
Advanced solo developers on GNOME Linux who need agent-driven control of native apps and panels for testing or personal automation.
Skip if: Builders on macOS or Windows, headless-only VPS workflows without a GUI session, or teams uncomfortable with privileged Docker desktop access.
What do I get? / Deliverables
Your agent can discover accessible UI nodes and send input on GNOME Wayland so desktop QA and automation scripts run from the chat.
- Structured AT-SPI UI snapshots for agent reasoning
- Scriptable desktop input actions via Mutter
- Repeatable agent workflows over native GNOME applications
Recommended MCP Servers
Journey fit
Desktop automation support belongs in Build when you extend agent tooling beyond the IDE to real Linux GUI apps you ship or dogfood. Agent-tooling is the canonical shelf because the server is infrastructure for agents to see and act on the GNOME desktop, not a product feature by itself.
How it compares
Linux GNOME desktop MCP automation, not Playwright-in-Chrome or a cross-platform mobile device farm.
Common Questions / FAQ
Who is GNOME UI MCP for?
Linux-focused developers using MCP agents who run GNOME on Wayland and need to automate or inspect native desktop UI.
When should I use GNOME UI MCP?
Use it when browser automation cannot reach your target UI and you need AT-SPI-backed discovery plus Mutter input from an agent.
How do I add GNOME UI MCP to my agent?
Pull ghcr.io/asattelmaier/gnome-ui-mcp:0.4.1, configure your MCP client to launch it via Docker with stdio, host network, and your host uid/gid as in the server schema, then reconnect the agent on the GNOME session.