Jumbo is in beta. Non-backward compatible changes could be introduced. Please post issues if you discover anything and contributions are welcome.

Jumbo - Memory for Coding Agents

Use Jumbo.
Focus on goals, not context.

Quick Start • Installation • Usage • FAQs

---

Can we address the elephant in the room?

Face it, working with coding agents is not ALL fun.
One step back for every two forward is not a productivity boost.
It's a hassle.

Context engineering alleviates the pain, but has you focused on the wrong things.

How does Jumbo help?

Jumbo is a CLI tool that gives your coding agent memory like an... well, 🐘.
It keeps your agent on track, and you focused on what you want to build.

Seen in Action

Before Jumbo:

Monday - Claude Code CLI:
You: 
   "I'm building an Event Driven distributed system..."
Claude: 
   *helps for 2 hours*

Tuesday - New chat:
You: 
   "Continue working on my Event Driven distributed system"
Claude: 
   "What Event Driven distributed system? Can you describe your project?"
You: 
   *spends 20 minutes re-explaining everything*

After Jumbo:

Monday - Claude Code CLI:
You: 
   "I'm building an Event Driven distributed system..."
Claude: 
   *helps for 2 hours*

Tuesday - New chat (or different AI tool):
You: 
   "Hi"  
Claude:
   "Hi! I can see we have 2 planned goals ready to work on:

     - Model the Auth domain
     - Implement Auth abstractions

    Would you like to start with one of these goals, or did you have something else in mind?"

Seen in Benefits

• Tracks project details
  Important aspects of you project are stored, retrievable, and mutable.
  _{(see below)}

• Delivers optimized context
  Agents receive the context they need to work aligned with YOUR criteria.

• Portable
  Switch models, or move from CLI to IDE, without interruption. Jumbo stays with you project and knows exactly where you left off.

• Extends context windows
  Optimal context delivery lets agents work longer.

• Automatic
  Hooks into your agent session and orchestrates the flow. It just works.

• Full control
  Jumbo's memories are yours. Stay in control and manage your data directly from the terminal.

• Private
  All data stays local. Nothing leaves your machine.

• Fast
  No network calls. No lag. Works instantaneously.

What's in the trunk?

Jumbo models memories as the following aggregates:

• Project
  What you are building and who it's for.
• Architecture
  Your solution design, structure and patterns applied.
• Components
  The parts comprising your solution and their roles.
• Dependencies
  What your project relies on.
• Decisions
  A history of why you chose what you chose.
• Guidelines
  The preferences, best practices, and the standards you adhere to.
• Invariants
  The rules you simply won't compromise on.
• Goals
  What you're working on—now, next and later.
• Sessions
  The state of the project.
• Context Packets
  Optimized context packets delivered to your AI agents.

---

Quick Start

# Install globally
npm install -g jumbo-cli

# Initialize in your project
jumbo init

That's it. Jumbo will guide you through the rest. Fire up your coding agent next and they'll use Jumbo automatically.

---

Documentation

Resource	Description
Quickstart	Get running in 5 minutes
Installation	Prerequisites and setup
Concepts	Understand sessions, goals, and context
Goal Management	Complete guide to tracking work
Command Reference	Full command documentation

---

Dependencies

Jumbo is built with:

Package	Purpose
better-sqlite3	Local event store and projections
commander	CLI framework
chalk	Terminal styling
yaml	Context serialization
inversify	Dependency injection
ulid	Time-sortable unique IDs

---

FAQs

How does jumbo integrate with my AI agent?

Through hooks. Your agent calls jumbo session start at the beginning of a session, and jumbo injects relevant project context. A richer context packet is delivered to the agent when it starts work on a goal. New insights are captured in the natural flow of your agent conversations.

What if I change agents or models?

Change agents and models at will. jumbo just picks up where you left off.

What coding agents does jumbo work with?

jumbo has been battle tested with Claude Code CLI, Gemini CLI, and Copilot CLI. More to be verified soon...

What IDEs are supported?

Theoretically, any IDE with an integrated coding agent should work. VS Code running GitHub Copilot has been tested and works well with all supported models. Cursor to be verified soon...

Where is data stored?

Locally, in .jumbo/ within your project. Nothing leaves your machine unless you want it to.

Can I control what data Jumbo captures?

Absolutely. You control how you want your agent to interact with Jumbo. Stay in-the-loop by approving each command, or run with pre-approved Jumbo commands for an automated experience.

Is Jumbo going to hijack my agent?

Not at all. Jumbo prescribes an opinionated workflow that you can always bypass. It works alongside your agent to enhance its capabilities.

Why not just use markdown files?

Jumbo goes beyond static markdown files. It's an immutable event stream—capturing your entire project history, always current and auditable. You stay in your flow, never repeat yourself—only add new information when you need to. Markdown is a snapshot in time, Jumbo is your project's living memory.

Can I share context across a team?

Not yet. A teams version is coming soon.

If you're feeling bold, you can try committing .jumbo/ directory to your repository—not recommended though. Without very tight coordination you're bound to encounter problems. jumbo uses Event Sourcing under the hood, working asynchronously will definitely result in out-of-sequence events.