Skip to main content
Coming in the first release. This page describes behaviour being built for the first Gliss release. Details may change before it ships.
Gliss works with the AI tool you already use: Claude Code, Claude Desktop, Cursor and Codex, or any tool that reads AGENTS.md and speaks MCP.

Start with one sentence

Tell your AI:
For AI agents is written for your AI. It checks the gliss client, signs you in, creates the app and builds it, and asks you when it needs you. The whole site is also available to AI tools as Markdown: add .md to any page’s address, or start from llms.txt at the root of this site.

The guidance in every app

Every Gliss app carries instructions for your AI, so it writes gliss.yaml correctly, keeps secrets out of the code and uses mock mode: gliss init writes them into a new app. In any app’s folder: Commit the files, so everyone working on the app gets the same guidance. Run gliss guide --write again after gliss update.

Connect the Gliss tools

gliss mcp gives your AI the Gliss tools: status, diagnosis and logs. It uses your gliss login session.

Letting your AI release

By default your AI can read status, diagnoses and logs, but not change anything. To let it release from your machine too, start the server with --allow-release (Beta and your Preview) or --allow-release=preview (your Preview only), and --dir for the app’s folder. It then asks you before every release. This builds with Docker on your machine, so it needs what gliss release needs. Soon most people won’t need it: each push to your work branch will build your Preview.
Never allowlist or auto-approve gliss_release. Each call replaces what runs at a stage.