March 12, 2026
Sharpened code, blunt memes
Show HN: Axe A 12MB binary that replaces your AI framework
Tiny 12MB “Axe” vows to chop bloated AI tools — devs cheer, nitpick, and meme the logo
TLDR: Axe is a tiny 12MB command-line app that chains small AI tasks like building blocks, aiming to replace bulky frameworks. The comments split between minimalism fans praising “run and done” workflows and skeptics missing chat-style sessions—plus a surprise meme war over whether the logo is a hammer or an axe.
The dev world is losing it over Axe, a tiny 12MB command-line app that says it can replace your sprawling AI framework by chaining small, focused “agents” like old-school Unix programs. Fans call it a return to sanity: no chatroom vibes, no “forever session,” just pipe in text, get results, move on. One commenter even compared it to ell, the cult-favorite tool for treating prompts like tiny programs.
But the first fight broke out over philosophy: Is a world without chat-style sessions actually better? When someone asked “There is no ‘session’ concept?”, the crowd split. Supporters say Axe keeps memory as simple timestamped notes and lets Unix handle scheduling (cron jobs, git hooks). Skeptics want the cozy comfort of a running chat. Another camp is buzzing about real-life uses, with one fan gushing that Axe matches their routine: fire once, check results later, especially with local models. Meanwhile, the meme patrol showed up hard: “Is the axe drawing actually a hammer?” sent the thread into chaos-emoji mode.
Under the hood, Axe speaks to OpenAI, Anthropic, and even local models via Ollama, all from TOML config files that you can version. The vibe? Minimalist rebels vs. framework faithful, with a side quest to rename the logo and a chorus yelling, “Cool idea—now show us what you’ve automated already.”
Key Points
- •Axe is a CLI tool that runs LLM-powered agents as small, composable programs defined in TOML files.
- •It supports multiple providers (Anthropic, OpenAI, Ollama) and integrates with Unix tools (cron, pipes, git hooks) without a daemon or GUI.
- •Features include sub-agent delegation, persistent memory with logs, memory garbage collection, a skill system, dry-run mode, JSON output, and sandboxed file/shell tools.
- •Installation requires Go 1.24+, with quick-start commands to initialize config, create/edit agents, run agents, and pipe input from tools; example agents are provided.
- •A Docker image enables isolated execution with multi-architecture builds; agents run with mounted configs and API keys, and Axe exits with code 2 if no config is present.