Axe is a lightweight command-line tool, inspired by the Unix philosophy, that allows you to define and run single-purpose agents based on large language models (LLMs) within a terminal environment. Unlike traditional, complex AI agent frameworks that are always running and consume significant computing resources, Axe provides a lightweight engine that is instantly executed only when needed, by combining a TOML-formatted agent configuration file and a set of Markdown directives called Skills, created by the user. Just as a Unix pipeline precisely processes and passes along a stream of text, Axe also smoothly connects the input and output of individual tools to structure complex artificial intelligence processing. Existing Python-based agent frameworks have been difficult to integrate with real-time responsive systems automation infrastructure, such as Git hooks or Cron jobs, due to their heavy library loading and slow boot times. Axe, in the form of a lightweight Go binary, is provided as a single executable file of approximately 12MB, allowing it to be launched instantly without additional dependencies, and transparently supports multiple cloud providers, from local Ollama to Anthropic and OpenAI. It also records timestamp-based Markdown memory to maintain context during execution and incorporates an intelligent garbage collection mechanism that compresses unnecessary patterns from accumulated conversation history using artificial intelligence, delivering optimal performance within a limited token budget. Typical researchers and software developers can use this tool to build various automation workflows concisely. For example, in the process of processing large-scale genomic analysis data, unstructured error log files can be piped as standard input (Stdin) to an analysis agent, and the agent's summarized results can be piped again to a messenger delivery tool. Furthermore, a multi-agent flow in which the agent automatically delegates tasks to other sub-agents can be fully coordinated solely through the relationships in local TOML configuration files.

Axe is a lightweight command-line tool that follows the Unix philosophy, allowing you to define and run single-purpose agents based on large language models (LLMs) in a terminal environment. Unlike existing complex AI agent frameworks that are always running and consume large amounts of computing resources, Axe combines a TOML-formatted agent configuration file written by the user and a set of Markdown directives called Skills, executing them only when needed.

Filtering Large-Scale Sequencing Error Logs: Establish a cat nextseqrun42error.log | axe run error-analyzer pipeline to perform real-time analysis of 50,000 lines of unstructured bioinformatics data logs using the Claude 3.5 Sonnet model, extracting key error causes and solutions within 3 seconds to control data loss rate at 0%.