Your AI can generate dashboards, forms, and tools — but the output is stuck in a chat window. Gui gives it a URL. One API call, instant shareable page, real-time multiplayer built in.
GUI turns HTML, Markdown or Mermaid into a live page in one call. Use whichever of these your agent already has — they all hit the same endpoint and return the same shareable link.
Typed tools — create_canvas, create_markdown_canvas, create_diagram, update_canvas. Typed arguments mean the HTML never passes through hand-written JSON escaping, which is where these calls usually break.
Zero dependencies. If your agent can run a shell command, it can publish a page:
npx -y gui-now push dashboard.html --title "Q4 Revenue"One POST, no account, no API key on the free tier:
curl -X POST https://gui.now/api/canvas \
-H 'Content-Type: application/json' \
-d '{"html": "<h1>Hello</h1>"}'POST the HTML to https://gui.now/api/canvas and you get back a live URL you can open or send to anyone. No account, no build step, no hosting. The page renders immediately, JavaScript runs, and forms work. Agents that support MCP can call the create_canvas tool instead, and agents with a shell can pipe a file through "npx -y gui-now push".
Install the GUI agent skill with "npx skills add gui-now/skills --skill gui-now -g". It works in Claude Code, Cursor, Codex, Windsurf and any agent that supports the Skills protocol. Once installed, the agent posts dashboards, charts and reports to a URL and shares the link instead of printing HTML into the conversation.
Yes. The GUI MCP server exposes create_canvas, create_markdown_canvas, create_multi_frame, create_diagram, update_canvas and extend_canvas as typed tools. Typed arguments matter here: passing a full HTML document inside a hand-escaped JSON string is where these calls usually break.
Free canvases stay live for 24 hours and accept 3 edits. Pro canvases last up to 30 days, accept unlimited edits, and can be password-protected. Updating a canvas pushes the change to everyone currently viewing it over server-sent events, so a shared link stays current without anyone reloading.
POST the diagram source to https://gui.now/api/flow and you get a URL showing a pannable, zoomable rendering. No Mermaid install and no local rendering step — useful when an agent produces a diagram that would otherwise arrive as unreadable source text.
Yes. Every canvas subscribes to a server-sent events stream, so an update pushed through the API appears for everyone already watching. Inputs, textareas and selects also sync between viewers, which makes a canvas usable as a shared scratchpad rather than a static page.