Open-source tools, infrastructure, and APIs for composable LLM behavior & adaptive intelligence.
Don't let your AI stagnate.
Take control with Bread.
&color=222&style=for-the-badge&logo=twitter&logoColor=white){kind=small}
Bread gives people—not oligopolies—control over AI.
Today’s language models are static, one-size-fits-all, and centralized. We’re building open tools to break that paradigm. Bread delivers cloud and open-source systems enabling precise, versioned, and composable updates—at the weight level—of LLMs. You can teach, refine, or update a model just by describing what you want, in natural language, as easily as correcting a friend.
Why?
- Machine learning shouldn’t be locked behind a PhD.
- You shouldn’t need specialist teams or corporate buy-in to adjust model behavior.
- Everyone should be able to build, adapt, and own intelligence that works their way.
We’re fundamentally against centralization of AI control, and we believe adaptive, iterative teaching—by demonstration, question/answer, and practice with feedback—makes models both more useful and more human-aligned.
No “AI priesthood.” No static black-boxes. Only intelligence you can teach.
-
⚡️ aibread SDK
Official JavaScript/TypeScript SDK to interact with our cloud API and run prompt bakes. Composable edits, dependency tracking, and more. -
🗃️ aibread·mcp
“Multipurpose Control Panel” microserver. User-friendly abstraction around the SDK for orchestrating model edits, chaining bakes, and collaborative workflows. -
🌱 aibread·bgit
Git-style “branch & merge” versioning for machine learning artifacts. Bring order and clarity to LLM weight changes. -
🐍 PyPI: aibread
Python tools for editing and composing model behaviors (early availability, more OSS tools “coming soon”).
- aibread.com
- PyPI: aibread
- npm: @aibread/sdk
- npm: @aibread/mcp
- npm: @aibread/bgit
- X (Twitter)
- Discord
- Composable LLM “baking”: Stack, chain, and recompose model edits as easily as merging branches.
- Natural language weight editing: Edit and update model behavior using instructions and feedback, no deep learning expertise needed.
- Open and modular: Connect with our cloud or (soon) self-host open tools for surgical model versioning.
- Decentralized control: Bread is fundamentally committed to empowering everyone to adapt, teach, and own AI.
💬 Want open-source libraries? More OSS tools are coming soon—stay tuned!
- 🔜 More open-source tools & self-hosted options (“coming soon”)
- Improved docs, guides, and walkthroughs (contributions welcome!)
- Community bakes, bake sharing, and collaborative workflows
We welcome researchers, engineers, and the merely AI-curious.
Check individual project repos for contribution guidelines—or join Discord to get involved.
“Never stop learning.”