Axonix Open Source: Why We Build in Public and the Projects Worth Cloning

We don't like walled gardens

The internet was built on open standards. HTTP, HTML, SMTP, RSS. These protocols are free, documented, and anyone can build on them. That openness is why the web became what it is.

Over the last decade, we've watched that shift. APIs require paid subscriptions. Platforms lock you into their ecosystem. Tools that could be open get wrapped in a SaaS pricing page.

As developers, this frustrates us. We like building things. We prefer building things that other people can use, modify, and improve without asking permission.

That's why we open-source our projects. Not as a marketing strategy. As a principle.

Project 1: PocketMCP

If you've been following the AI space, you've heard of the Model Context Protocol. MCP is a standard that lets AI agents interface with real-world data sources. Instead of an AI being confined to a chat window, it can read files, query databases, send messages, and interact with external systems.

Most MCP servers are desktop-bound. They run on your computer as Python scripts or Node processes. They're useful, but they're limited to what your desktop can access.

Your phone can do things your desktop can't. It has GPS. It has SMS. It has a camera. It has cellular connectivity. It's always with you.

PocketMCP turns any Android device into an MCP server. It's a native Kotlin application that runs on your phone and exposes your phone's capabilities to AI agents on your desktop over your local network.

Here's what that means in practice:

You're sitting at your desk working with an AI assistant. You ask it to send a text message to your colleague. Instead of using a paid API like Twilio, the AI sends the request to PocketMCP on your phone, and your phone sends the SMS directly. No third party. No per-message cost.

You ask the AI to calculate your commute time. It pulls your phone's current GPS location, checks the route, and gives you an accurate estimate based on where you actually are.

You ask the AI to take a photo and analyze it. Your phone's camera becomes an input for your desktop AI session.

The project bridges the gap between the intelligence on your desktop and the physical capabilities in your pocket. It's useful for developers building AI agents that need real-world interaction. It's useful for anyone who wants their phone to be more than just a screen.

How it works technically

PocketMCP runs as a background service on Android. It exposes an MCP-compatible API over your local network. Your desktop AI client discovers it automatically and connects. All communication stays on your local network unless you explicitly configure remote access.

The codebase is Kotlin with Jetpack Compose for the UI. It uses WebSocket for real-time communication and implements the full MCP specification for tool discovery, invocation, and result handling.

Why we open-sourced it

We could have kept PocketMCP closed and built a subscription service around it. The market for AI-to-phone integration is growing. But the developer community makes these tools better. When someone submits a pull request that fixes a race condition in the WebSocket implementation, everyone benefits. When another developer adds support for a new phone capability, the whole ecosystem gets stronger.

Project 2: ClearBox

Your email inbox is a mess. Mine is too. We built ClearBox because we were tired of it.

ClearBox is an open-source React and Next.js dashboard that connects to your Gmail account through the Gmail API. It's not another email client. It's a search-and-destroy tool for the things cluttering your inbox: marketing spam, cold outreach, forgotten newsletter subscriptions, and promotional emails you never asked for.

Here's what it does differently:

It analyzes your inbox. ClearBox scans your unread emails and categorizes them by type. Marketing. Social notifications. Financial receipts. Travel confirmations. You see a breakdown of what's actually in your inbox instead of a wall of unread messages.

It integrates with local AI. We connected ClearBox to Ollama for on-device AI processing. The AI analyzes your emails and provides summaries, identifies patterns, and suggests actions. Because it runs locally through Ollama, your email content never leaves your machine.

It gives you bulk actions. Found 47 unread marketing emails from the same sender? One click unsubscribes you from all of them. Found 200 promotional emails you'll never read? Bulk delete them by category.

It roasts you. This is the fun part. ClearBox will tell you that you're holding onto 4,000 unread marketing emails from high-performing Buy and that maybe, just maybe, it's time to let go. It's not mean about it. But it's not gentle either.

How it works technically

ClearBox uses the Gmail API for read and write access to your inbox. The OAuth flow gives it permission to read, categorize, and delete emails. The AI processing runs through Ollama on your local machine, so email content is analyzed locally. The dashboard is a Next.js application with a clean interface for reviewing categories and taking bulk actions.

Why we open-sourced it

Email tools are a crowded market. There are dozens of inbox cleanup apps, most of them paid. We built ClearBox because we needed it ourselves. We open-sourced it because other people need it too, and the community can make it better than we can alone.

When someone adds support for Outlook. When another developer improves the AI categorization. When a contributor builds a mobile version. Everybody wins.

Why we open source everything

If we kept these projects closed-source, we could monetize them. PocketMCP could be a five-dollar-a-month subscription. ClearBox could be a high-quality app.

But the reality is that open source makes these tools better. Faster. More reliable. More useful to more people.

When someone finds a bug we missed and submits a fix, the tool improves. When someone adds a feature we hadn't thought of, the tool grows. When someone forks the project and builds something we never imagined, the ecosystem expands.

This isn't altruism. It's pragmatism. Open source is the fastest way to build good software.

How to get involved

Don't just star the repositories. Fork them. Use them. Break them. Submit issues when something doesn't work. Submit pull requests when you fix something.

The code is on GitHub. The documentation is in the repos. The issues are open. If you have an idea for a feature, open a discussion. If you find a bug, open an issue. If you want to contribute code, open a pull request.

Grab the projects from our Open Source page or directly from GitHub. Help us build tools that are actually useful.

Frequently asked questions

Are these projects free to use?

Yes. Both PocketMCP and ClearBox are open source under permissive licenses. You can use them for personal or commercial projects without paying anything.

Do I need to self-host these tools?

PocketMCP runs on your Android device. ClearBox can be deployed to any hosting platform that supports Next.js, or run locally on your machine. Both are designed to be self-hosted.

Can I contribute to these projects?

Absolutely. We welcome contributions of all kinds: bug fixes, new features, documentation improvements, and bug reports. Check the CONTRIBUTING.md file in each repository for guidelines.

Are these projects production-ready?

PocketMCP is stable for local network use. ClearBox is functional for personal inbox management. Both are actively maintained and improving. Check the GitHub repositories for the latest status and known issues.

What happens if I find a security issue?

Open a private security report through GitHub's security advisory feature. We take security seriously and will respond promptly. Both projects handle sensitive data (SMS, email), so security is a top priority.

Final note

Open source is how we believe software should be built. In public. With community input. Without gatekeeping. If that resonates with you, check out PocketMCP and ClearBox. Clone them. Use them. Make them better.