smallnest/pigo

pi agent in Golang

What it solves

pigo is a command-line AI coding agent designed to automate the cycle of understanding requirements and modifying code. It bridges the gap between high-level prompts and concrete file changes by allowing an LLM to read, write, and execute commands directly within a local development environment.

How it works

The project implements a layered runtime architecture where a central agent loop manages the interaction between the user, the LLM provider, and a set of local tools. It uses a two-layer loop system: an inner loop that handles the stream of responses and tool executions until the assistant stops calling tools, and an outer loop that manages follow-up messages.

Key operational components include:

  • Multi-Provider Support: Compatible with OpenRouter, Ollama, Anthropic, and OpenAI-compatible endpoints.
  • Tool Integration: Built-in tools for file operations (read, write, edit), search (grep, find), shell execution (bash), and web access (webfetch, websearch).
  • Security Gates: A project trust system that requires user confirmation for "side-effect" tools (like bash or write) in untrusted directories.
  • Context Management: Automatic context compression (summarization) when approaching the LLM's window limit.

Who it’s for

  • Developers who want an AI agent capable of performing actual coding tasks in their terminal rather than just generating snippets.
  • CI/CD Engineers who can use the "headless" mode (-p) for automated scripts.
  • Go Developers looking to embed coding agent capabilities into their own applications via the provided SDK.

Highlights

  • Dual Execution Modes: Supports both an interactive REPL and a headless script mode for CI/CD integration.
  • Extensible Architecture: Features a system for "Skills" (reusable prompt-based capabilities), plugins, and a robust Hooks system to intercept or modify agent behavior via shell commands.
  • Flexible Tool Control: Granular tool access via white-lists (--allowed-tools) and black-lists (--disallowed-tools).
  • Session Persistence: Ability to list, resume, and continue previous sessions.
  • Prompt Templating: Supports reusable Markdown templates with complex parameter syntax for common tasks.

Related

  • Project
    cline/cline

    An open-source coding agent that integrates into IDEs and terminals to autonomously edit code, run bash commands, and manage complex development workflows.

  • Project
    go-vgo/robotgo

    A Golang library for desktop automation that enables programmatic control of the mouse, keyboard, and screen for auto-testing and AI computer use.

  • Project
    tbckr/sgpt

    A Go-based command-line interface for interacting with OpenAI and other compatible LLMs to generate shell commands, code, and text responses directly in the terminal.

  • Project
    MigoXLab/dingo

    Dingo is an open‑source Python library (and SaaS product) for evaluating AI data, model outputs, and RAG systems. It supports rule‑based, LLM‑based, and agent‑driven checks, streams data from files, databases, S3, or Hugging Face, and provides JSON reports plus optional visual dashboards. Install via `pip install dingo-python` (with optional extras), run via CLI (`dingo eval …`) or Python SDK, and extend with custom rules, LLM prompts, or agents.

  • Project
    aandrew-me/tgpt

    A cross-platform command-line interface tool that allows users to interact with AI models directly from their terminal.