Good work starts with shared context.

For AI-first teams: builders, product managers, and the people deciding what comes next.

Get startedTry the demo

How unbranch works: one canonical model, every change agreed.

How unbranch works

One canonical model.
Every change agreed.

unbranch is the concept architect for your product. Your team and every AI tool it uses work from one canonical model: the single source of truth for what you are building and why.

Three separate threads join one shared notebook.

The whole picture

Three roles, the tools they already use, one canonical model.

Your team

Product managers, developers, and everyone deciding what comes next.

Your tools

already in use

PM

Claude & ChatGPT

Talk an idea through. The assistant writes the draft.

Developer

Coding agent

Reads the model before building, checks the work after.

Everyone

The portal

Draft a change, see every proposal, give consent.

MCP server

The one way AI tools reach unbranch. They can read and draft. They never decide.

Read the model

Draft a change

Check the work

AI audit

AI

Flags duplicates and conflicts. Never decides.

Draft

draft

A change, written down.

Proposal

in review

Sent to everyone to answer.

Canonical model

source of truth

Where the product is going, what people can do, and what it won't do.

Directions

Capabilities

Constraints

Consent

people

Every member answers: consent, abstain or object.

Commitment

agreed

Not started → in delivery → verified.

Ledger

Every change, on the record.

Notifications

Reviews, objections, deliveries.

  • Your team to Your tools: works in
  • Claude & ChatGPT to MCP server: asks · drafts
  • MCP server to Draft: writes the draft
  • The portal to Draft: drafts a change
  • Draft to AI audit: checked by AI
  • Draft to Proposal: submitted
  • Proposal to Consent: to everyone
  • Coding agent to MCP server: reads · checks the work
  • MCP server to Canonical model: reads the model
  • The portal to Consent: consent · abstain · object
  • Consent to Commitment: accepted
  • Commitment to Canonical model: becomes true
  • Consent to Ledger: on the record
  • Ledger to Notifications: every change
  • Notifications to Your team: keeps everyone posted

Try it your way

Bring your AI tools,
or open the portal.

Three ways in, and none of them comes first. Each works on its own, and they all meet in the same canonical model.

PM

Developer

With your AI tools.

Claude on the web or Claude Code in your repository. Both come in through unbranch's MCP server: they read the model and draft a change, and never decide.

claude.ai

Connectors

Settings

Add custom connector
Add unbranch to Claude

Everyone

In the portal.

Where the whole team sees the model and every proposal, and where each person answers. Only people can.

portal.unbranch.ai

Lumen Studio / Lumen / Canonical model

Canonical modelas of ledger entry #40
MapTree
LiveArchive
DirectionDelivered
Readers save from anywhere
2 capabilities
Stack them
CapabilityDelivered
Save an article from any page
CapabilityActive
Share sheet on phones
DirectionActive
Reading that keeps your place
CapabilityDelivered
Pick up where you stopped
DirectionDelivered
A calm place to read
2 capabilities
Stack them
CapabilityDelivered
Reader view without ads
CapabilityDelivered
Adjustable type and spacing
DirectionActive
Highlights you can find again
CapabilityActive
Highlights sync across devices
ProposalSave articles to read offline
raised by Maya Chen
1 capability1 constraintIn review
Owner
MCMaya Chen · edited 16 hours ago
Versions
r1Save articles to read offline · 1 version
Your response
ConsentAbstainObject
Record response
Where it lands in the model
DirectionActive
Reading that keeps your place
lives under
CapabilityProposed
Save an article to read offline
ConstraintProposed
Saved articles stay on the device
Proposed — not in the model yet
Try the demoGet started

Let’s stay in touch.

Have a question or an idea to share?

Send us feedback in a short form, or email us at info@unbranch.ai.

Share feedback