Quick Start

Prerequisites

• A Remb account (free tier available)
• macOS, Linux, or Windows with Homebrew / Go installed

1. Install the CLI

brew install samie105/remb/remb

# Latest release at https://useremb.com/download

Verify the installation:

remb --version

2. Authenticate

remb login

This opens your browser for OAuth. Once approved, credentials are stored locally at ~/.config/remb/credentials. Alternatively, set the REMB_API_KEY environment variable.

3. Initialize a Project

cd your-project
remb init

Creates a .remb.yml config and registers the project. You can link a GitHub repo now or later — it's only needed if you want automated scanning.

No GitHub repo? Use remb__project_init_local from your AI via MCP, or skip linking entirely — Remb is fully useful without a scan.

4. Connect Your AI Client

Run the local MCP proxy so your AI tools can read and write context:

remb serve

Then point your AI client at it — see the MCP overview for VS Code, Cursor, Claude Code, and Windsurf config snippets.

Install skills so your AI knows how to use Remb automatically:

remb skills add --all

5. Start Engineering Context

Remb is a context engineering platform — your project's context is something you build and curate, not something that gets extracted once from source files. Start immediately:

# Save a feature note
remb save -f auth "Uses RS256 JWT, tokens expire in 15m, refresh in Redis"

# Save a memory
remb memory add --title "Deploy pattern" --content "Always run db:migrate before deploy" --category pattern

# Your AI calls these via MCP:
remb__memory_create   # save a decision or pattern
remb__context_save    # document a feature
remb__conversation_log  # record what was accomplished

Every session your AI works in, it adds to the context graph. By the next session it already knows your architecture, decisions, and open problems — without re-reading the codebase.

Scanning (optional)

If you have a GitHub repo connected, you can trigger an automated scan to bootstrap context from your codebase structure — useful for large existing codebases where manually documenting every feature would take time.

remb scan       # trigger cloud scan, poll progress
remb push       # same + git status pre-check

Scanning extracts features, tech stack, and file relationships — but the context your AI builds through use (memories, decisions, conversation logs) is equally valuable and starts working immediately.

Next Steps