Providers

A provider is the agent behind a conversation. OAR runs several of them side by side, each with your own account or key, and you pick one when you start a conversation. These pages cover what each provider needs, how to switch it on, and what works differently there.

• OverviewThe agents OAR runs side by side, where each one runs and what it needs, a table of what each can do, and how to install a provider's CLI from the relay.
• GitHub CopilotRun GitHub Copilot in OAR - what you need, the SDK and Extension engines, the Copilot settings, usage, and what to do when it fails.
• OpenAIRun an agent on your own OpenAI API key, or chat with OpenAI's image models: settings, Quality and Size, Edit this image, and what removing the key does.
• ClaudeRun Claude on your own machine through OAR - the login it needs, the Claude settings, the CLI row, commit attribution, effort and thinking controls, and fixes.
• Claude APIRun Claude chats on a Claude Console API key instead of your subscription - the key, switching a chat between the subscription and the key, the monthly cap, the usage card, models and efforts, what stays apart, error codes and fixes.
• Claude CloudRun Claude Code in a sandbox at Anthropic on a GitHub repository, driven from OAR - switching it on, starting a cloud chat, what it cannot do, and fixes.
• CursorRun Cursor's agent on your own machine through OAR - the API key, the Cursor settings, models and effort, what each OAR feature does on Cursor, usage figures and fixes.
• GrokRun the Grok CLI through OAR - installing it from the relay, switching Grok on, signing in with a device code, models, what works and what does not, usage and fixes.
• DeepSeekRun DeepSeek's models through OAR - the API key, the models and their efforts, how a DeepSeek chat runs Claude Code on the relay's machine, web search, the usage card with its balance, budget and charts, the usage export import, how the estimate is made, error codes and fixes.
• Models and reasoningChoose the model and its reasoning effort: the pickers, Select Models, where each model list comes from, context sizes, Claude's thinking, Claude API's list and DeepSeek's levels.