Skip to content

Latest commit

 

History

History
35 lines (20 loc) · 2.11 KB

File metadata and controls

35 lines (20 loc) · 2.11 KB

Comparisons

opencode-llm-proxy is deliberately narrow: it reuses the providers you already configured in OpenCode and exposes them through common client API formats. Here's how it relates to adjacent tools.

vs LiteLLM

LiteLLM is a general-purpose provider router/proxy. opencode-llm-proxy is focused on reusing the providers already configured in OpenCode and making them available through common client API formats.

  • Use LiteLLM if you want a standalone production gateway with its own provider configuration, routing, budgets, and key management.
  • Use opencode-llm-proxy if your model configuration already lives in OpenCode and you want local tools to reuse it without re-entering keys.

vs OpenRouter

OpenRouter is a hosted router and provider marketplace with its own billing and API keys. opencode-llm-proxy is local and backed by your OpenCode setup.

  • Use OpenRouter for a hosted, internet-facing endpoint and access to a broad model catalog through one account.
  • Use opencode-llm-proxy to keep everything local and driven by the providers OpenCode already has authenticated.

vs direct provider API keys

Configuring a provider's API key directly in each tool is simplest when you only have one tool.

  • Use direct keys for a single tool with a single provider.
  • Use opencode-llm-proxy when multiple tools should share the same OpenCode model setup, so you configure providers once and swap models in one place.

vs Ollama's OpenAI-compatible endpoint

Ollama exposes an OpenAI-compatible API for its local models only.

  • Use Ollama's endpoint if you only ever use local Ollama models.
  • Use opencode-llm-proxy to expose Ollama and every other provider OpenCode manages (GitHub Copilot, Anthropic, Bedrock, OpenRouter, ...) behind one endpoint and multiple API formats.

Summary

opencode-llm-proxy is not trying to be a production gateway or a hosted marketplace. Its value is that your model setup already lives in OpenCode, and this makes that setup usable from OpenAI-, Anthropic-, Gemini-, and Responses-API-compatible tools locally, with streaming and real tool/function calling.