open-source coding agents

NVIDIA has released the code of its agent safety platform as open source, making its tooling for keeping AI agents secure and controlled available to developers. Technology publications including Open Source For You and Adafruit's blog report the move, which lets companies inspect and adapt the safety software rather than rely on a proprietary product.

A new open-source tool called claude-mem, published by developer thedotmack, is gaining attention on GitHub. Written in TypeScript, it captures everything an AI coding agent does during a session, compresses it with AI, and reinjects relevant context into future sessions. It is designed to work with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode and other agents.

A developer has released an open-source tool on Hacker News that routes requests between AI models for coding agents, claiming performance comparable to Astra, a reference to high-tier results on coding benchmarks. The launch drew engagement from the developer community, with discussion centring on whether a routing layer over existing models can match single frontier models on coding tasks.

Google engineer Addy Osmani has published agent-skills, an open-source JavaScript repository offering production-grade engineering skills for AI coding agents. The project is trending on GitHub, drawing attention from developers interested in making AI assistants follow real-world engineering practices rather than ad-hoc prompts.

Trigger.dev has introduced Plimsoll, an open-source Python sandbox that lets AI agents safely execute code while analysing data. Agents often need to run calculations, inspect files, or test snippets, and the sandbox provides an isolated environment for that. Developers in the open-source community are sharing the tool as a practical addition to agent-building workflows.

Developer Julius Brussee released Caveman, an open-source Go project that acts as a proxy and skill for AI coding agents, rewriting their communication into terse 'caveman' speech. The project claims this cuts token usage by roughly 65%, lowering costs for tools like coding assistants. Its joke framing, riffing on a well-known meme phrasing, has helped it gain attention online.

Cloudflare has published an open-source security-audit skill for AI coding agents, available on GitHub under the name security-audit-skill. Written in JavaScript, the tool runs multi-phase security audits and produces machine-readable findings that are independently verified, aiming to make automated code review more trustworthy. Developers are picking it up and discussing how it could fit into agent-based development workflows.

A developer has released Graphene, an open-source data analysis toolkit designed to give coding agents better tools for working with data. The project, available on GitHub, is being shared with the Hacker News community and drawing attention from people following the fast-moving space of AI coding assistants and their ecosystems.

Developer Corey Haines released an open-source repository called marketingskills that gives Claude Code and other AI agents a set of marketing capabilities, including conversion rate optimization, copywriting, SEO, analytics and growth engineering. Written in JavaScript, the project is drawing attention among developers and marketers exploring how to turn coding assistants into growth tools.

A developer has released Sidequest, an open-source Kanban board described as one where 'the cards work themselves', pairing project management with AI-driven automation. Writing on Eigenwise, the author shares August usage figures from running the tool in real software development work. Early attention is coming from the open-source and AI coding communities, where interest centres on how agents might automate task execution rather than just tracking it.

Agency Agents is an open-source platform offering a collection of AI agent personas designed to work like a digital agency, handling tasks such as coding, development and engineering. The project, which claims more than 146,000 stars on its repository, is being discussed in developer and AI communities, with supporters highlighting its inclusive, community-driven approach to building software agents.

NanoMuse, an open-source AI agent designed to run on both mobile phones and computers, has been released on GitHub. The project is drawing attention in the developer community, with early engagement suggesting interest in a self-hosted, cross-device AI assistant as an alternative to closed commercial agents. Details on its features and capabilities remain limited beyond the code repository.

Anthropic's Claude Code, an open-source agentic coding tool written in TypeScript, is gaining traction among developers. The tool runs in the terminal, understands a codebase, and executes tasks like routine edits, code explanations, and git workflows through natural language commands. Its rising visibility suggests growing interest in AI assistants that work directly inside developers' existing workflows rather than separate chat interfaces.

The MLC team has published TIRx Harness, an open-source compiler harness designed to let AI agents write and optimize GPU code. The project combines compiler infrastructure with agentic workflows, aiming to automate low-level GPU programming tasks that normally require deep expertise. Reaction so far is limited but positive, with early readers on Hacker News discussing the potential for agents to handle performance-critical kernel work.

A developer says his coding workflow no longer requires him sitting at his desk: he now writes code from his phone, with the open-source AI coding agent Cline doing the typing inside VS Code. He credits the setup with removing the hard requirement of being physically present, calling Cline a great driver while he steers from a distance. The post has drawn attention in software and productivity circles, where people are weighing how far AI agents can take over hands-on development work.

A new writeup examines how developers can build stateful voice AI agents using dynamic graphs with the Felona framework, an open-source approach written in TypeScript. The piece argues that intelligent voice agents have long attracted developers and businesses, and walks through techniques for managing conversation state. It is being shared within coding and open-source communities interested in voice AI development.

The Manyfold 3D printing project, an open-source model library application, is being discussed after its AGENTS.md file — instructions written for AI coding agents working on the repository — was shared and praised as refreshingly sensible. Commenters appear to be highlighting it as a good example of how projects can set expectations for automated contributors.

A developer has built an open-source Telegram bot that uses Google's Gemma model running locally to read and understand bills, then sends persistent reminders through Temporal's durable workflow system until the bill is paid. Sentry is used for agent tracing without exposing bill data. The project is part of a weekend coding challenge and is being shared openly with the developer community.

A security researcher has released Tanuki, a new open-source tool described as protocol-first Linux Active Directory triage built for AI coding agents and operators. The developer says most coding agents hallucinate or propose noisy, dangerous tactics when handling Linux AD environments, such as recommending weak crypto settings. The project is being shared with the infosec community.