Show HN: Char – Ship a UI-first copilot (no backend rewrite)
We built an embedded agent that acts through your existing UI permissions via the WebMCP standard.
We built Char because we were tired of watching teams rewrite their entire applications just to add a “copilot.”
The problem: Most agents are built backend-first. They require you to rebuild your business logic as “tools,” re-implement your permission system, and then struggle to handle simple things like “are you sure?” modals or client-side validation.
The solution: WebMCP. It’s a standard we’re proposing (and dogfooding) that brings the Model Context Protocol to the browser.
- Your UI is the API: Use
useWebMCP()hooks to expose tools directly from your components. - UI Parity: The agent sees what the user sees. If the user can’t click “Delete,” neither can the agent.
- One-Shot Setup: Because it’s all standard MCP, you can use a coding agent (like Claude Code) to scaffold the tools, test them in the browser, and embed the agent in about an hour.
Enterprise Unlock: Cross-App Interop (FDC3 for Agents)
The most interesting part? Because the agent lives in the browser session, it can bridge multiple internal apps.
Imagine an agent that can read a ticket in your Support App and trigger a refund in your Billing App, without you needing to build a complex backend mesh. It’s like FDC3 (the finance desktop standard), but for AI actions.
We are looking for Design Partners
We are looking for 3–5 teams who have multiple internal apps and want to unify them with a copilot.
If you are a “Ford-shaped” organization (lots of internal web apps, real governance needs), we want to build with you.