Docs · Quickstart

Two commands
to see it work.

Unfurl runs on Node.js. During early access you install it from source; the steps below are the same ones in the README.

1 · Try it with no credentials

Seed a messy dataset

The demo plants eight common data problems, then resolves them. doctor shows what was caught.

# build, then seed a messy sample dataset (no credentials)
$ npm install && npm run build
$ node dist/cli.js demo
# see what the resolution engine caught
$ node dist/cli.js doctor
2 · Connect your agent

Add it to Claude Code

Any MCP client works the same way: point it at unfurl serve over stdio. Then ask "what is my open pipeline by stage?", "who churned recently?", or "which won deals never started paying us?"

# plug it into Claude Code
$ claude mcp add unfurl -- node /path/to/unfurl/dist/cli.js serve
3 · Use real data

Sync HubSpot and Stripe

Create a HubSpot private app with read scopes for contacts, companies, deals, and owners. Optionally add a Stripe restricted key, read-only on customers and subscriptions.

$ node dist/cli.js init
# writes .unfurl.json; add hubspotToken (and optionally stripeKey)
# or set HUBSPOT_TOKEN / STRIPE_KEY, which override the file
$ node dist/cli.js sync
4 · Look around

Open the local dashboard

The same resolved graph, in your browser, served from localhost only.

$ node dist/cli.js ui
# http://127.0.0.1:4173
Reference

Every command

unfurl init      unfurl demo      unfurl sync      unfurl ui
unfurl resolve   unfurl doctor    unfurl serve
Early access

Get the code.

Request early access and we will send you the repository and help you connect your first source.

Get early accessSee the tools