Swival/swival

A small, powerful, open-source CLI coding agent that works with open models.

What it solves

Swival is a coding agent designed to work across different AI models, whether they are hosted locally or via online services. It addresses the challenge of maintaining context and security when switching between models of varying sizes and capacities.

How it works

Swival operates as a Python-based tool that allows users to set goals for coding tasks. It manages conversation space and working notes to help smaller models perform effectively. For security, it includes secret encryption to hide API keys and tokens, and filters to remove private details like customer names and internal URLs before sending data to the model.

Who it’s for

Developers who want a flexible coding agent that can switch between local and online models without changing the agent's logic, and those who need built-in security audits and review steps for their code.

Highlights

  • Model Agnostic: Works with both local and online AI models.
  • Security Focused: Includes secret encryption and private detail filters.
  • Context Management: Maintains working notes to track progress on long-term goals and resume interrupted sessions.
  • Built-in Review: Supports a review step where the agent or a separate model can check for bugs and errors.
  • Extensible: Written in Python and can be integrated into other tools or run as part of a Python program.

Related

  • Project
    inferstep/ATLAS

    A local coding agent that enhances compact open models with planning, sandboxed verification, and energy-based quality scoring to perform software engineering tasks on consumer hardware.

  • Project
    qwen-code-dev-bot/oh-my-cli

    A secure, self-hosted code agent CLI that reads and edits files and runs shell commands using OpenAI-compatible endpoints with a focus on safety and session durability.

  • Project
    Doorman11991/smallcode

    A terminal-native coding agent optimized for local LLMs (8B-35B parameters) that uses intelligent context management and tool routing to enable reliable coding on consumer hardware.

  • Project
    Sentdex/minion

    A lightweight coding agent designed to eliminate context bloat, enabling efficient operation on local LLMs and remote APIs while providing tools to read, write, and execute shell commands.

  • Project
    zqiren/Orbital

    A project-centric agent coordinator that stores project context as local files, allowing users to switch between different AI agents without re-explaining the project.