Axe is a CLI tool that reimagines AI agents as small, focused utilities rather than monolithic chatbots. Instead of building another framework that assumes you want a long-running conversation with massive context, Axe treats LLM agents like Unix programs - each doing one thing well and composable with standard tools.

Built in Go, Axe orchestrates agents defined in TOML configuration files with their own system prompts, model selection, skill files, and persistent memory. The tool supports multiple providers including Anthropic, OpenAI, and Ollama, allowing users to pipe data directly into agents or chain them together. Agents can delegate to sub-agents, maintain context across runs, and be triggered from cron jobs, git hooks, or CI pipelines.

Installation requires Go 1.24+ and involves a simple binary with minimal dependencies. The tool includes features like dry-run mode for inspecting context, JSON output for scripting, and built-in file operations all sandboxed to the agent's working directory. Docker support provides isolated, hardened containers with non-root users and read-only filesystems by default, making Axe suitable for both development and production environments.