Today's GitHub trending is a clear snapshot of where developer attention is concentrated right now: coding agents. From free-access wrappers to Rust-native terminals to single-file configuration guides, the community is rapidly assembling the infrastructure that makes AI-assisted development more accessible, more customizable, and more embedded in everyday workflows. The big theme is democratization — both of access and of know-how.
Free-Claude-Code: Community Access to AI Coding Tools
Alishahryar1/free-claude-code is trending with nearly 50,000 stars, offering a way to use Claude Code alongside Codex, Pi, and OpenCode from the terminal, web app, or IDE — advertising access to over 1.3 billion free tokens in the process. Projects like this follow a familiar open-source pattern: when a commercial tool generates enormous developer interest, the community builds accessibility layers almost immediately. The speed of adoption here says as much about appetite for agentic coding tools as it does about price sensitivity.
github.com/Alishahryar1/free-claude-code
OpenAI Codex: A Terminal Coding Agent Written in Rust
openai/codex carries over 117,000 stars and is trending again today — a sign that developers are still discovering and sharing it rather than it being a one-day spike. Written in Rust, it's a lightweight coding agent designed to run directly in your terminal. Compared to heavier IDE-integrated tools, Codex's terminal-first approach appeals to developers who prefer staying in their shell and composing agentic behavior with standard Unix primitives. The Rust choice signals a clear focus on performance and cross-platform portability.
Andrej Karpathy's Claude Code Skills: A Single File, 200K Stars
multica-ai/andrej-karpathy-skills is a striking entry: one CLAUDE.md file — the instruction format Claude Code reads before starting a session — compiled from Andrej Karpathy's public observations about how to get more reliable behavior from the tool. With over 206,000 stars, it's one of the most-starred projects trending today, which is remarkable for a single-file repository. It reflects how much developer interest there is in tuning agentic behavior through instructions and prompts rather than code changes.
github.com/multica-ai/andrej-karpathy-skills
NousResearch Hermes Agent: An Agent That Grows With You
NousResearch/hermes-agent tops the star count on today's list at over 236,000, described simply as "the agent that grows with you." NousResearch is known for fine-tuned open models, and hermes-agent appears to be a framework for building persistent agents that accumulate context and capabilities over time — a contrast to the stateless, single-session interactions most agent frameworks assume. The deliberately minimal description is generating curiosity clicks, with developers clearly intrigued by what an agent that "grows" actually delivers in practice.
github.com/NousResearch/hermes-agent
Apache Maka: Local-First AI Agent Workspace
apache/maka is a TypeScript project from the Apache Software Foundation billed as a "local-first AI agent workspace." Rather than relying on a database or cloud backend, it records all model messages and tool interactions as append-only logs — the same architectural pattern used in event sourcing and distributed systems. For developers who care about auditability, offline reliability, and keeping data under their control, this is a meaningful design departure from cloud-dependent agent frameworks, and it's gaining traction fast with over 3,000 stars today.