henrikdukefoss

Agent Provider Configs — AI skill for Claude Code

AI community

Copy-paste configs to point Claude Code, OpenCode, Codex CLI, Cline, Kilo, Aider, Hermes, OpenClaw, Pi, Zed, Cursor and 10 more AI coding agents at any OpenAI- or Anthropic-compatible API.

How to install Agent Provider Configs

This entry records only its repository, not the path inside it, so there is no exact command to give. Open henrikdukefoss/agent-provider-configs and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Provider Configs does

Copy-paste configs to point Claude Code, OpenCode, Codex CLI, Cline, Kilo, Aider, Hermes, OpenClaw, Pi, Zed, Cursor and 10 more AI coding agents at any OpenAI- or Anthropic-compatible API. Sourced gotchas and provider base URLs.

Alternatives in AI

LLM For Zotero — An open-sourced research agent system deeply rooted in your Zotero library 2.8k ★
Rulesync — by dyoshikawa - A Node.js CLI tool that automatically generates configs (rules, ignore files, MCP servers, com 909 ★
Coffee CLI — A clean AI terminal workspace 180 ★

README

agent-provider-configs

Copy-paste configs for pointing AI coding agents at any OpenAI-compatible or Anthropic-compatible API: a gateway, a router, another model vendor, or a server on your own machine.

20 agents, 21 provider presets, a source link for every key. Last checked: **2026-10-14**.

How do I point an AI coding agent at a custom OpenAI-compatible API?

You need three values: a **base URL**, an **API key**, and a **model ID** that the provider serves. Where they go depends on the agent:

Claude Code reads env vars: ANTHROPIC_BASE_URL (no /v1), ANTHROPIC_AUTH_TOKEN, ANTHROPIC_MODEL. It only speaks the Anthropic Messages API.
OpenCode and Kilo Code take a provider block in opencode.json / kilo.json with "npm": "@ai-sdk/openai-compatible" and options.baseURL.
Codex CLI takes a [model_providers.] table in ~/.codex/config.toml. It only speaks the OpenAI Responses API now; wire_api = "chat" was removed.
Cline, Roo Code, Cursor, Copilot, JetBrains have an "OpenAI Compatible" option in their settings UI.
Most other agents take a JSON or YAML provider entry. Find yours in the table, copy the block, replace the three placeholders.

Placeholders used everywhere: `https://YOUR_PROVIDER/v1`, `YOUR_PROVIDER_API_KEY`, `MODEL_ID`. Real values for 21 providers are in [Provider presets](#provider-presets).

The four mistakes behind most "404" and "model not found" errors

/v1 or no /v1. OpenAI-style base URLs end in /v1 (the client appends /chat/completions). Claude Code, Zed's anthropic_compatible, Goose, Roo Code and Factory's anthropic type want the Anthropic base URL without /v1 because they append /v1/messages themselves. OpenCode, Kilo, Continue and Crush use AI SDK or similar clients that want it with /v1. Each section below says which.
Wrong protocol. Codex CLI needs /v1/responses. Claude Code needs /v1/messages. An endpoint that only serves `/v1/chat/c