Getting started
One scan maps every feature from git history, scores it, and serves it to your team and your AI agent. Here's how to install it, run it, and wire it into your editor.
How it works
Wystan is two pieces that share one artifact. The engine scans a repo and writes a feature map; the hosted MCP server serves that map to your AI agent, and the same map drives PR comments and the weekly Slack digest. Connect a repo, we run the first scan, point your editor at the MCP — and your agent stops guessing about your code.
Connect the MCP
The MCP is a hosted, OAuth-secured server — nothing to install. Claude Code, Claude Desktop, Cursor (1.0+), VS Code (1.101+) and Codex CLI all speak remote MCP natively. In Claude Code:
claude mcp add --transport http wystan https://app.wystan.ai/api/mcpThen run /mcp inside Claude Code to authorise — the first call opens
your browser to sign in. In Cursor or Claude Desktop, add a custom connector
with the same URL. Clients that only speak stdio can bridge through
npx -y mcp-remote https://app.wystan.ai/api/mcp (a community
proof-of-concept — use the native path where you can).
Run a scan
Connect a GitHub repository in the dashboard (read-only access) and we run the first full scan for you — the working copy is read in memory and discarded after the run. The result is the feature map: product features, the user flows inside them, and the files and symbols each one touches, each scored.
After every push an incremental scan re-maps only what changed, so the map — and your agent's context — never goes stale; PR branches get their own current scan, separate from main. Prefer to run the engine yourself? It is source-available (FSL-1.1); the hosted flow above is the supported path today.
Connect your agent
Any MCP-compatible client works — Cursor, Claude Code, Cline, Aider,
Continue. Drop the server into the client's MCP config and restart it.
For Cursor that's ~/.cursor/mcp.json:
{
"mcpServers": {
"wystan": { "url": "https://app.wystan.ai/api/mcp" }
}
}Then ask your agent “what features touch checkout?” or “what's the regression risk of this diff?” and it answers from the feature map instead of grepping blind.
Deployment modes
The tool API is designed to be identical across three deployment modes — only the data path changes. Hosted is what ships today; the other two are on the roadmap and listed here so you can see where the design goes.
The MCP server runs in our cloud; your agent calls it over HTTPS with an org-scoped token. Lowest setup friction, and the only mode that works with hosted agents like Claude.ai web. Sentry + PostHog joined server-side.
The server would run as a local process on the developer's machine, pull encrypted scans from us, decrypt with a key that never leaves the machine, and serve the agent data we never see in plaintext. Not available today.
The entire stack — engine, dashboard, MCP — would deploy inside your infrastructure, with agents calling your internal endpoint. On the roadmap; if you need it, tell us: hello@wystan.ai.