Local-first by design
Your projects and application data live on your machine by default. You choose how Gaggle connects to AI providers.
YOUR WORKSPACE. YOUR AGENT. YOUR RULES.
An AI agent that works where you do.
Gaggle is an open-source, local-first AI assistant built to work with your files, projects, and development workflows — right on your machine.
01 / The idea
AI should fit your workflow. Not the other way around.
Working with AI should not require you to rethink how you work. Your projects already have context, and your tools already have a place in your workflow. Gaggle explores a more focused approach: bringing assistance closer to the environment where your work happens.
02 / What matters
Everything starts with the context that matters.
Six design principles keep the assistant close to the work, without crowding it out.
Built around your projects
Work in real directories, explore files, and keep your AI sessions connected to the projects that matter.
Permission-aware execution
Define how tools operate and require approval for sensitive actions, with enforcement handled by the application.
A workflow developers understand
Connect AI-assisted work with familiar development tasks, including file inspection, terminal operations, and Git review where supported.
Flexible AI integrations
Connect supported AI providers through a modular integration layer, without tying the entire application to one provider.
Open source, built to evolve
Explore the code, understand how it works, and help shape the future of a developer-focused AI assistant.
03 / The experience
More than a conversation.
From understanding your project to carrying out real work, Gaggle helps turn instructions into meaningful actions. This interface preview shows a workspace-centered flow: selected context, visible progress, and next steps you can review without losing your place.
Conceptual interface preview — illustrating the ideas behind a workspace-centered approach to AI assistance, not a screenshot of a finished product.
See how it works Task progressToday, 10:42 AM
Read workspace contextatlas-web / package.json
0.4sReview navigation patterns3 files inspected
1.2sPrepare next stepsWorking with selected context
•••The agent keeps you in the loop, without getting in the way.
04 / Privacy & control
Your machine. Your projects. Your choice.
AI should work with your workflow without taking control away from you. Gaggle is designed to keep application data local by default while giving you control over provider connections and tool permissions.
Local data by default
Project metadata, conversations, and execution history are designed to be stored locally.
You choose your AI provider
When using a remote provider, requests and relevant context may leave your device. Understand what is sent and how the provider processes it.
Actions with guardrails
Use permission checks and explicit approvals for sensitive operations, backed by enforcement in the application.
Visibility into agent activity
Make it easy to inspect tool activity, execution status, and proposed changes.
05 / The workflow
From your workspace to meaningful work.
Bring your project
Open a local workspace and give Gaggle the context it needs to help with your work.
Describe what you want to do
Explain a problem, explore an idea, or ask for help with a development task in natural language.
Review and move forward
Inspect the results, review proposed changes, and approve actions when required.
06 / Built for developers
Thoughtful engineering. Open by design.
Gaggle is an open-source project built around local workflows, thoughtful engineering, and a community-driven future.
Local application runtime
A Rust desktop application with SQLite local persistence, with macOS and Linux in mind. The focus is a foundation that remains approachable as it grows.
Agent and tool execution
Built-in tools for filesystem, shell, memory, web, and browser access, plus MCP extensions, all routed through a permission engine.
Provider integrations
OpenAI-compatible model providers are supported through a modular adapter layer, with credentials stored in the OS keychain.
Rust SQLite Axum OpenAI-compatible macOS + Linux
07 / Get started
Get started with Gaggle.
Explore the source, set up your environment, and follow the installation guide for your platform.
Gaggle is built from source today. You will need a recent Rust toolchain and Git.
Prerequisites
Rust toolchain supporting the 2024 edition (rustc 1.85+) and Git.
Build
git clone https://github.com/frengkysorituamanurung/gaggle.git cd gaggle cargo build --workspace
Run the server
cargo run -p gaggle-cli -- serve --addr 127.0.0.1:4321
Then open http://127.0.0.1:4321 in your browser.
08 / Built in the open
The future of Gaggle is built together.
Good tools get better when the people who use them can help shape them. Explore the code, share your ideas, report issues, and contribute to the next chapter of Gaggle.
09 / FAQ
Questions, answered.
What is Gaggle?
Gaggle is an open-source, local-first AI assistant designed to work with your files, projects, and development workflows.
Does Gaggle run locally?
Gaggle is designed to run its application services and store application data locally. Using a remote AI provider can still send prompts and relevant context to that provider. Local model support depends on the integrations available in your version.
Which platforms does Gaggle support?
Gaggle targets macOS and Linux. Check the current installation documentation for the platforms and architectures verified for your release.
Do I need an AI API key?
That depends on the provider you choose. Remote providers may require their own API credentials, while local model integrations may have different requirements.
Is Gaggle open source?
Visit the Gaggle repository to review its license, source code, contribution guidelines, and current development status.
Make your workspace work harder.
Meet an AI assistant built around your projects, your workflow, and your control.