docs / panels / Info

Info

The Info button on the main toolbar — the just after Help — opens a floating window with three things in it: the diagnostics of the session you are running right now, a button that starts the MCP server so an AI agent can drive this window, and the licences this application is distributed under.

Its neighbour, Help, opens this help site.

Licences

The window carries the project licence and the maintained third-party notices inside the binary, so they are readable in a desktop build that was never served beside its help site. They are the same files this site renders under Licence and Third-party notices.

Open the third-party crate inventory opens a page of this site instead. That page is generated from the dependency graph every time the site is built, so it lists what the application actually ships.

Diagnostics

Row	What it tells you
Renderer	`WebGPU` or `WebGL2` in a browser; `Vulkan`, `Metal`, `Direct3D 12` or `OpenGL` on the desktop.
wgpu backend	The same backend under the name the graphics layer gives it, for matching against its own documentation.
Adapter	The graphics device that is drawing — a GPU's model name, or a software rasteriser's.
Device type	Discrete GPU, integrated GPU, or CPU. A CPU device explains a great deal of slowness on its own.
Driver	The driver name and version, where the backend reports one. Browsers usually do not.
Max texture size	The largest texture edge this graphics device supports — the ceiling on how large the 3D view can be drawn.
App version	The build you are running.
Platform	The browser build, or the desktop build and its operating system.

Renderer is the row most worth knowing. The browser build asks for WebGPU and falls back to WebGL2 when the browser cannot provide it; which one you got changes what the 3D view can do and how fast it is, and it is not something the picture on screen tells you.

These are read once, at startup, from the device that is actually drawing — not measured again when the window opens.

Connect an agent (MCP)

Start MCP server starts the embedded MCP server inside the window you have open — the same thing launching the application with --mcp does, without the flag. Click it once; the button then reads MCP server running and stays disabled, and the section fills with the text to paste into your agent — the very text --mcp prints to the terminal, selectable, with a Copy agent instructions button under it:

BREP MCP server listening on http://127.0.0.1:8765/mcp
This window is the session: an agent's tools act on the documents you see here.
Every request needs the server's token, `Authorization: Bearer <token>`; the token is in
  ~/.config/brep-app/mcp-token (readable by you only).

Claude Code:
  claude mcp add --transport http brep http://127.0.0.1:8765/mcp --header "Authorization: Bearer $(cat ~/.config/brep-app/mcp-token)"
  or in a project's .mcp.json, with BREP_MCP_TOKEN set to the token:
    { "mcpServers": { "brep": { "type": "http", "url": "http://127.0.0.1:8765/mcp", "headers": { "Authorization": "Bearer ${BREP_MCP_TOKEN}" } } } }

Codex, with BREP_MCP_TOKEN set to the token (export BREP_MCP_TOKEN=$(cat ~/.config/brep-app/mcp-token)):
  codex mcp add brep --url http://127.0.0.1:8765/mcp --bearer-token-env-var BREP_MCP_TOKEN
  or in ~/.codex/config.toml:
    [mcp_servers.brep]
    url = "http://127.0.0.1:8765/mcp"
    bearer_token_env_var = "BREP_MCP_TOKEN"

Then ask the agent to use run_script; screenshots and call logs land under
  /tmp/brep-mcp

Run the line for your agent once; from then on an agent started while this window is open sees the brep tools and works on the documents you are looking at. Screenshots and call logs go under the OS temporary directory in brep-mcp/, as the last lines say.

The button listens on port 8765 like the flag. If that port is taken — another copy of the application already has it, say — it tries the next few ports up to 8774 and says which one it got, in a line above the text; the text itself always names the port actually in use, so what you copy is right. If none of them is free, or the server cannot start for another reason, the line the terminal would print appears under the button instead, and you can try again.

If you launched with --mcp, the server is already running: the section shows the same text with the live address and the button is disabled — there is one server per window, never a second. A build without the MCP server (the browser build, or a desktop build made without it) says so under the button.

The same block travels with a bug report

Copy diagnostics puts the block on the clipboard for a report you are filing elsewhere. You do not need it for the in-app one: the Submit Bug button attaches exactly this block to every report, and shows it to you before it is sent.

Nothing in the block identifies you. It is hardware and build facts, and the report's description — which is where the block travels — may be shown publicly on the problem list.