Orchestrate multi-repo workspaces
with scoped permissions for AI agents
Define your workspace topology once. Let AI understand where it can—and can't—make changes.
npm install -g binder-cli What is Binder?
Binder is a CLI tool for managing multi-repository workspaces with AI assistants.
When you work across multiple repos, AI doesn't know which ones you own vs. which you're just referencing.
Binder solves this by generating a CONTEXT_MAP.md file
that gives AI agents a semantic map of your workspace with explicit permissions.
Three Scopes. Clear Boundaries.
Binder uses three permission levels to tell AI exactly what it can do in each repository.
OWNER
Your domain. Full authority.
- Full read/write access
- Refactor freely
- Make breaking changes
PLATFORM
Guest contributor. Tread carefully.
- Limited write access
- Follow existing patterns
- Minimal changes only
REFERENCE
Read-only. For context.
- No modifications
- Read for understanding
- Never suggest changes
How It Works
Define
Create a binder.yaml that maps your workspace topology.
name: my-workspace
repos:
- name: my-service
url: git@github.com:org/my-service.git
scope: owner
- name: shared-platform
url: git@github.com:org/platform.git
scope: platform
- name: legacy-api
url: git@github.com:org/legacy-api.git
scope: reference Pull
Binder clones all repositories idempotently.
$ binder pull
✔ Cloned my-service (owner)
✔ Cloned shared-platform (platform)
✔ Cloned legacy-api (reference) Generate Context
Generate CONTEXT_MAP.md for your AI assistant.
$ binder context
✔ Generated CONTEXT_MAP.md What AI Sees
The generated CONTEXT_MAP.md gives AI clear instructions about each repository.
my-service at ./repos/my-service shared-platform at ./repos/shared-platform legacy-api at ./repos/legacy-api Quick Start
# Install
npm install -g binder-cli
# Initialize workspace
binder init
# Edit binder.yaml to add your repos...
# Clone everything
binder pull
# Generate AI context
binder context
# Check status anytime
binder status Works With Your AI Tools
Binder generates markdown context that works with any AI tool that accepts system prompts or context files.