LLMSwap documentation

LLMSwap provides one direct interface to 10 LLM providers through a Python SDK, CLI, and optional local browser Arena.

Start here

• Getting started: install, configure, diagnose, and ask
• CLI reference: Ask, Best Answer, Arena, MCP, providers, and workspaces
• Provider setup: credentials and current default models
• Python SDK: queries, Best Answer, async use, tools, and controls

Core workflow

                 ┌─> Ask one provider
Prompt -> LLMSwap├─> Compare selected models in the Arena
                 └─> Synthesize a cross-checked Best Answer

LLMSwap uses your provider accounts. Remote prompts go to the providers you select; Ollama can remain local when the Ollama server is local.

Best Answer privacy model

Prompt -> 2–5 candidate requests in parallel
       -> anonymized candidate text + prompt -> judge
       -> answer + agreement + disagreements + cautions

The default uses one provider/model for candidates and judge. Cross-provider output sharing requires explicit consent. LLMSwap does not automatically redact sensitive data, and agreement does not guarantee factual correctness.

Additional references

• Provider and model audit
• Security policy
• Examples
• Changelog

---

Table of contents

• Getting Started
• CLI Reference
• Provider Setup
• Python SDK