A developer released Marmel 0.9.0, an autonomous coding agent designed to build prototypes, after two weeks of stability and reliability work focused on local models. The agent runs on small models like Gemma 4 12B and cloud models like DeepSeek v4/v4.1 Flash, and the author is seeking community testing, feedback, and pull requests via GitHub.
Marmel is a personal multiagent orchestration framework that coordinates local LLMs to perform long-running coding tasks. Its creator used it to autonomously build a simple x64/Linux C compiler from scratch over about six weeks. It targets the creator's own setup and is not yet easily configurable for others.
Marmel is a personal multiagent orchestration framework that coordinates local LLMs to perform long-running coding tasks. Its creator used it to autonomously build a simple x64/Linux C compiler from scratch over about six weeks. It targets the creator's own setup and is not yet easily configurable for others.
After a month of talking with MCP developers, a developer built and open-sourced MCPfy, a platform aimed at becoming the one-stop open source solution for building software used by AI agents. The project is early-stage and invites community feedback and contributions.
After a month of talking with MCP developers, a developer built and open-sourced MCPfy, a platform aimed at becoming the one-stop open source solution for building software used by AI agents. The project is early-stage and invites community feedback and contributions.
A developer is building open-agent, an open-source, self-hosted, BYOK alternative to the Grok bot, featuring a real agent loop, browser control, and memory. All CI is gated, and the developer invites others to test and contribute on GitHub. It matters to practitioners seeking self-hosted agent stacks they control end to end.
A developer is building open-agent, an open-source, self-hosted, BYOK alternative to the Grok bot, featuring a real agent loop, browser control, and memory. All CI is gated, and the developer invites others to test and contribute on GitHub. It matters to practitioners seeking self-hosted agent stacks they control end to end.
Claw-coder is a fully autonomous local coding agent that runs entirely on the user's machine, addressing privacy concerns by preventing data leaks to external servers. It uses local AI models to perform coding tasks safely, with built-in safeguards against dangerous actions. Targeted at developers who need a private, secure alternative to cloud-based coding assistants.
Claw-coder is a fully autonomous local coding agent that runs entirely on the user's machine, addressing privacy concerns by preventing data leaks to external servers. It uses local AI models to perform coding tasks safely, with built-in safeguards against dangerous actions. Targeted at developers who need a private, secure alternative to cloud-based coding assistants.
Autobuilder is an AI-assisted system designed to self-replicate and improve itself, aiming to perform work on a large scale. It uses LLMs to automate tasks and enhance its own capabilities, targeting the problem of increasing data center resource demands. The project is open-source and seeks community involvement for development.
Autobuilder is an AI-assisted system designed to self-replicate and improve itself, aiming to perform work on a large scale. It uses LLMs to automate tasks and enhance its own capabilities, targeting the problem of increasing data center resource demands. The project is open-source and seeks community involvement for development.