How AI Coding Agents Could Make Free Software Matter Again

For decades, the free software movement felt like a noble but fading cause. SaaS won because convenience beat ideology. But AI coding agents are about to change that calculation entirely—making software freedom practical for everyone, not just programmers.

The SaaS Convenience Trap

Richard Stallman founded the Free Software Foundation in 1985 with a simple principle: users deserve the freedom to run, study, modify, and share the software they depend on. These "four freedoms" powered the open-source revolution—Linux, Apache, MySQL, PHP—but then SaaS arrived and made them feel irrelevant.

The problem? When software runs on someone else's servers, having the source code doesn't help you. You can't modify Salesforce or Google Docs even if you had their code. The practical question became convenience, not freedom. No installation, automatic updates, someone else handling security patches and 3 AM outages.

But this convenience came at a cost: dependency. When your software doesn't work exactly how you need it to, you're stuck submitting feature requests and waiting years.

My Sunsama Workflow Nightmare

Here's a concrete example. I use Sunsama for task management and wanted a simple workflow: when I see a tweet I want to engage with later, save it as a properly categorized task. Sounds simple, right?

With an AI coding agent, this should be a twenty-minute project. The architecture is straightforward: catch the shared URL, extract tweet content, use an LLM to generate a smart task title, categorize it, and create the task. But Sunsama has no official API—and their feature request has been open since 2019.

To build this workaround, I needed:

• A reverse-engineered unofficial API (thanks to a community member)
• My actual Sunsama password stored in plaintext
• A serverless function I have to deploy and maintain
• A manually-built iOS Shortcut that can't be version-controlled
• Dependencies that could break any time Sunsama updates

Six layers of workarounds, three authentication mechanisms, infrastructure I'm now responsible for—all because the software isn't free.

Agents Bridge the Expertise Gap

Here's what changes with AI agents. Stallman's four freedoms always presupposed the ability to read and modify source code. For most users, that ability doesn't exist. As Protesilaos Stavrou argued, a free software license alone doesn't empower users if they lack the expertise to exercise those freedoms.

Agents flip this entirely. An AI coding agent is an intermediary that can exercise technical freedom on behalf of a non-technical user. When you tell Claude or Codex "make my task manager automatically categorize tweets," you're exercising Freedom 1—the freedom to study and modify—through a proxy.

You don't need to understand GraphQL schemas or React internals. You just describe what you want, and the agent exercises the technical freedoms for you.

The Self-Hosting Reality Check

Before you rush to self-host everything, consider the costs. Self-hosting means you're responsible for security updates, backups, SSL certificates, DNS, and operational overhead. There's also a sustainability concern: a 2026 working paper argues that AI-generated contributions can damage open source by severing user-maintainer feedback loops.

Adam Wathan, creator of Tailwind CSS, reported documentation traffic dropped ~40% even as usage grew, with revenue down ~80%. Mitchell Hashimoto moved Ghostty to a vouch-based contribution model in response to low-quality AI-generated PRs.

The answer isn't simply "run everything yourself." SaaS solved real problems. What we need are new models that combine the customization benefits of free software with SaaS convenience.

The New Buying Criterion

Over the next 1-2 years, expect this question to become standard: "Can my agent fully customize this?" It'll join "does this have a mobile app?" and "does it integrate with Slack?" as a core evaluation criterion.

John Gilmore famously said the Net interprets censorship as damage and routes around it. We're about to see the same for closed software. As agents become the primary way people interact with tools, they'll interpret unfree software as damage—an obstacle between users and what they want—and route around it.

Not because people suddenly care about software freedom ideology, but because they want their agents to actually help them. And agents can't help when software is locked down.

FAQ

What's the difference between free software and open source?

Free software (Stallman's definition) is about user rights and freedoms. Open source is a development methodology focused on code sharing. The 1998 "open source" rebrand kept the practices but removed the ethical claims about what users deserve.

Why did free software fade in importance?

SaaS created a licensing loophole. The GPL requires sharing source code with anyone you "distribute" software to—but if you never distribute it, just run it on servers, the license doesn't apply. Users never touched the code anyway, so the four freedoms became theoretical.

Can't I just use APIs instead of modifying source code?

Vendor APIs are limited by what the company chooses to expose. Many SaaS products have no API, or severely rate-limited ones. With free software, the entire codebase is available—no restrictions, no begging for features.

Will agents make self-hosting easier?

Yes, by lowering maintenance costs. Unlike the engineer who built your internal tool and left, agents don't leave. But the operational burden remains real.