Browse docs
--- title: "OpenCode and Other Agents" description: "Step-by-step guide for OpenAI- and Anthropic-compatible coding agents with Aurora." icon: "bot" --- Most coding agents outside Claude Code are easiest to connect to Aurora through the standard OpenAI- or Anthropic-compatible API. Flow: OpenCode / Cline / Roo Code / Kilo Code -> Aurora -> your upstream providers ## Before you start - Choose an Aurora master key, for example change-me. - Start Aurora with at least one upstream provider key. - Use Aurora standard http://your-aurora-host/v1 unless your client specifically requires a full endpoint URL. ## 1. Run Aurora Start Aurora with a master key and at least one upstream provider: ``bash docker run --rm -p 8080:8080 \ -e AURORA_MASTER_KEY="change-me" \ -e OPENAI_API_KEY="$OPENAI_API_KEY" \ aurorallm/aurora ` <Note> You can swap OPENAI_API_KEY for another provider key if that is what your target models use. The important part is that Aurora has at least one upstream provider configured. </Note> ## 2. Confirm the API with curl `bash curl -s http://your-aurora-host/v1/chat/completions \ -H "Authorization: Bearer change-me" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-4.1-mini", "messages": [ { "role": "user", "content": "Reply with exactly ok" } ] }' ` If the gateway is wired correctly, the response will contain ok. ## 3. Configure your coding agent ### OpenCode Create an opencode.json file with an Aurora provider: `json { "$schema": "https://opencode.ai/config.json", "provider": { "aurora": { "npm": "@ai-sdk/openai-compatible", "name": "Aurora", "options": { "baseURL": "http://your-aurora-host/v1", "apiKey": "{env:OPENAI_API_KEY}" }, "models": { "gpt-4.1-mini": { "name": "GPT-4.1 Mini", "limit": { "context": 200000, "output": 32768 } } } } } } ` Then export the Aurora key: `bash export OPENAI_API_KEY=change-me ` ### Cline In Cline, choose: - API Provider: OpenAI Compatible - Base URL: http://your-aurora-host/v1 - API Key: change-me - Model ID: a model returned by GET /v1/models ### Roo Code In Roo Code, choose: - API Provider: OpenAI Compatible - Base URL: http://your-aurora-host/v1 - API Key: change-me - Model ID: a model returned by GET /v1/models <Note> Roo Code requires native tool calling. Choose a model and provider path that support OpenAI-compatible tool calling. </Note> ### Kilo Code In Kilo Code, choose: - API Provider: OpenAI Compatible - Base URL: http://your-aurora-host/v1 - API Key: change-me - Model ID: a model returned by GET /v1/models Kilo Code also supports a full endpoint URL if you need it: `text http://your-aurora-host/v1/chat/completions ` Start with the simpler base URL first and only switch to a full endpoint URL if your setup requires it. ## 4. Run a test prompt For OpenCode, this validated command worked: `bash opencode run -m aurora/gpt-4.1-mini \ 'Reply with exactly ok and no punctuation.' ` The validated result was: `text ok ` For Cline, Roo Code, and Kilo Code, save the provider settings above and send a small prompt such as: `text Reply with exactly ok and no punctuation. ` ## 5. Check the traffic in Aurora Open the Aurora dashboard audit logs: http://your-aurora-host/admin/dashboard/audit This is the quickest way to confirm that your coding agent is reaching Aurora and to inspect each interaction. From the same dashboard, you can keep following your Aurora traffic and usage. ## Current status - OpenCode was validated locally and worked against standard http://your-aurora-host/v1 - Cline, Roo Code, and Kilo Code follow the same OpenAI-compatible setup model - Roo Code needs native tool calling support from the selected model - Kilo Code can use either a base URL or a full endpoint URL ## References - OpenCode providers docs: Providers - Cline OpenAI-compatible provider docs: OpenAI Compatible - Roo Code OpenAI-compatible provider docs: OpenAI Compatible - Kilo Code OpenAI-compatible provider docs: OpenAI Compatible ## Validated on March 10, 2026 This guide combines: - local validation against an Aurora instance on http://your-aurora-host - local validation with OpenCode 1.1.36 - official documentation for Cline, Roo Code, and Kilo Code Local validation confirmed: - OpenCode worked against http://your-aurora-host/v1 - the validated OpenCode request used POST /v1/chat/completions` - Aurora dashboard audit logs showed the OpenCode traffic immediately