Dev ToolsCode sourceFree

Swival

Open-source CLI coding agent with A2A and MCP support that works across many model providers.

Pricing

Free

Dev Tools

Curated resource

Technical badges

CustomPython1.0-alpha

Marketplace signals

Views2K
Gets0
RatingNo ratings yet

What it does

Long-form description, mapped skills, and a concrete usage example.

Swival is a practical CLI coding agent designed to work reliably even with smaller or local models. It supports A2A and MCP, connects to LM Studio, Hugging Face, OpenRouter, Google Gemini, ChatGPT, and generic OpenAI-compatible servers, and is designed for constrained-context coding workflows.

Autonomous code editing

Run a coding-agent loop against a local or hosted model and act on repository tasks.

Cross-provider A2A interoperability

Use A2A and MCP while targeting local models or popular hosted APIs.

Technical

Framework

Custom

Language

Python

Spec version

1.0-alpha

Interaction

Text, Streaming

Supported input

TextJSON

Supported output

TextJSON

MCP servers used

Local or external MCP servers

Requirements

API keys

OpenAI, Google AI, Anthropic

Infra

Python 3.11+

Estimated run cost

Runs free with local models; hosted model cost depends on provider.

External services

LM Studio, Hugging Face Inference API, OpenRouter

How to install

Clone or access the repository at https://github.com/Swival/swival.

Checkout the branch or tag `main`.

Prepare the required runtime: Python 3.11+.

Provision the required API keys: OpenAI, Google AI, Anthropic.

Connect the external services used by the agent: LM Studio, Hugging Face Inference API, OpenRouter.

Install dependencies, configure env vars, and deploy to your preferred host.

Install with `uv tool install swival` or Homebrew, then point it at LM Studio, Hugging Face, Gemini, ChatGPT, OpenRouter, or any OpenAI-compatible runtime.

Ratings

No ratings yet

0 total ratings
Ratings appear here after buyers receive their secure feedback link.

Comments

Buyer discussion

0 total comments
No comments yet. The first useful buyer note can set the tone here.

Leave a comment

Sign in with GitHub or Google to comment on agents.

Purchased already? Use the secure link from your confirmation email to rate or report this agent. The public page stays open for browsing, but feedback actions require a verified buyer or downloader token.