Providers

opencode supports 75+ LLM providers through the AI SDK and Models.dev integration. This includes major cloud AI platforms, open-source model hosts, local inference engines, and custom OpenAI-compatible endpoints.

Supported Providers

Provider	Description
Anthropic	Claude model family (Sonnet, Haiku, Opus)
OpenAI	GPT-4, GPT-5, o3, o4-mini models
Google Vertex AI	Gemini models via Google Cloud
Amazon Bedrock	Claude, Llama, Mistral via AWS
Azure OpenAI	OpenAI models hosted on Azure
Groq	Low-latency inference for open models
DeepSeek	DeepSeek-V2, DeepSeek-R1 models
Ollama	Local inference for open-source models
OpenRouter	Unified API for 200+ models
Together AI	Cloud inference for open models
xAI	Grok models
GitHub Copilot	Copilot-powered models
GitLab Duo	GitLab's AI features
Hugging Face	Hugging Face Inference Endpoints and Serverless
Deep Infra	Serverless GPU inference
Fireworks AI	Fast inference for open models
Cerebras	Wafer-scale AI acceleration
LM Studio	Local model server with OpenAI-compatible API
llama.cpp	Lightweight local LLM inference
Atomic Chat	Local chat interface with model serving
Mistral AI	Mistral models (Large, Small, Codestral)
Cohere	Command-R, Command-R+ models
Perplexity	PPLX models for search-augmented generation
Replicate	Cloud API for open-source models
Anyscale	Ray-based model serving
AI21 Labs	Jamba, Jurassic-2 models
Writer	Palmyra model family
OctoAI	Cloud inference platform
Lepton AI	Serverless model serving
Modal	Cloud functions for AI workloads
Noon AI	Korean language model provider
Nvidia NIM	GPU-optimized model containers
Sambanova	Dataflow architecture for LLMs
AI/ML API	OpenAI-compatible API

And many more via Models.dev. To see the full list:

opencode providers list

Adding Credentials

The easiest way to add provider credentials is through the TUI using the /connect command:

/connect

This opens an interactive prompt where you select a provider and follow the authentication flow. Credentials are stored securely in:

~/.local/share/opencode/auth.json

Authentication Methods

Providers support different authentication mechanisms:

API Key (most common):

/connect anthropic

Then paste your API key when prompted.

Environment Variables:

Set well-known environment variables for automatic detection:

export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
export GROQ_API_KEY=gsk_...
export TOGETHER_API_KEY=...

OAuth:

Some providers (GitHub Copilot, GitLab Duo) use OAuth flows. The /connect command opens a browser window for authorization.

Provider Configuration

Provider-specific options are set under the provider key in opencode.json.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "anthropic": {
      "options": {
        "baseURL": "https://api.anthropic.com/v1",
        "timeout": 600000,
        "chunkTimeout": 30000,
        "setCacheKey": true
      }
    },
    "openai": {
      "options": {
        "baseURL": "https://api.openai.com/v1",
        "timeout": 600000
      }
    }
  }
}

Provider Options Reference

Option	Type	Default	Description
`baseURL`	string	Provider default	API endpoint base URL
`timeout`	number	`600000`	Request timeout in milliseconds
`chunkTimeout`	number	`30000`	Streaming chunk timeout in milliseconds
`setCacheKey`	boolean	`false`	Enable prompt caching (Anthropic)
`reasoningEffort`	string	provider default	Reasoning budget (`none`, `low`, `medium`, `high`)
`thinking`	object	`null`	Extended thinking configuration
`textVerbosity`	string	`"normal"`	Response verbosity level

Provider-Specific Options

Some providers have unique configuration options.

Amazon Bedrock:

{
  "provider": {
    "amazon-bedrock": {
      "options": {
        "region": "us-west-2",
        "profile": "my-aws-profile",
        "endpoint": "https://bedrock-runtime.us-west-2.amazonaws.com"
      }
    }
  }
}

Google Vertex AI:

{
  "provider": {
    "google-vertex": {
      "options": {
        "projectId": "my-gcp-project",
        "location": "us-central1"
      }
    }
  }
}

Azure OpenAI:

{
  "provider": {
    "azure-openai": {
      "options": {
        "resourceName": "my-resource",
        "deploymentId": "gpt-4-deployment",
        "apiVersion": "2024-10-01-preview"
      }
    }
  }
}

Custom Providers

Any OpenAI-compatible API endpoint can be added as a custom provider using the @ai-sdk/openai-compatible package.

{
  "provider": {
    "my-custom-provider": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "My Internal Provider",
      "options": {
        "baseURL": "https://api.internal.company.com/v1",
        "headers": {
          "X-API-Key": "{env:INTERNAL_API_KEY}",
          "X-Organization": "engineering"
        }
      },
      "models": {
        "internal-llama-3": {
          "name": "llama-3-70b-instruct"
        },
        "internal-mixtral": {
          "name": "mixtral-8x22b-instruct"
        }
      }
    }
  }
}

Custom Provider Fields

Field	Required	Description
`npm`	Yes	The npm package implementing the provider interface
`name`	No	Human-readable display name in the TUI
`options`	No	Provider options including `baseURL`, `headers`, `timeout`
`models`	Yes	Map of model IDs to model definitions

Custom Headers

Use custom headers for authentication or routing to internal proxy services:

{
  "provider": {
    "internal-proxy": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Internal Proxy",
      "options": {
        "baseURL": "https://ai-gateway.company.com/v1",
        "headers": {
          "Authorization": "Bearer {env:GATEWAY_TOKEN}",
          "X-Tenant-ID": "engineering"
        }
      }
    }
  }
}

Local Model Providers

Local inference engines can be configured as OpenAI-compatible providers by pointing baseURL to the local server.

Ollama

{
  "provider": {
    "ollama": {
      "options": {
        "baseURL": "http://localhost:11434/v1"
      },
      "models": {
        "llama3.2": { "name": "llama3.2" },
        "mistral": { "name": "mistral" },
        "codellama": { "name": "codellama" }
      }
    }
  }
}

LM Studio

{
  "provider": {
    "lm-studio": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio",
      "options": {
        "baseURL": "http://localhost:1234/v1"
      },
      "models": {
        "local-model": { "name": "local-model" }
      }
    }
  }
}

llama.cpp

{
  "provider": {
    "llamacpp": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "llama.cpp",
      "options": {
        "baseURL": "http://localhost:8080/v1"
      }
    }
  }
}

Atomic Chat

Atomic Chat provides a built-in local inference server. Configure it like any OpenAI-compatible endpoint:

{
  "provider": {
    "atomic": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Atomic Chat",
      "options": {
        "baseURL": "http://localhost:1337/v1"
      }
    }
  }
}

Base URL Customization

Customizing baseURL is useful for:

Internal proxy or gateway services
Load balancers that distribute across provider regions
Air-gapped environments with a local API mirror
Provider compatibility layers (e.g., running Anthropic models through a proxy)

{
  "provider": {
    "anthropic": {
      "options": {
        "baseURL": "https://internal-ai-gateway.company.com/anthropic/v1"
      }
    }
  }
}

Troubleshooting

Check Stored Credentials

List all configured authentication entries:

opencode auth list

This shows which providers have credentials stored and when they were added.

Verify Provider ID

Provider IDs in the config must match the IDs used by opencode. Run the following to see exact provider IDs:

opencode providers list

Common provider IDs:

Config Key	Provider ID
`anthropic`	anthropic
`openai`	openai
`google-vertex`	google-vertex
`amazon-bedrock`	amazon-bedrock
`azure-openai`	azure-openai
`groq`	groq
`deepseek`	deepseek
`ollama`	ollama
`openrouter`	openrouter

Testing Provider Connectivity

opencode models anthropic

If the models list fails to load, check:

Network connectivity to the provider
API key validity
Base URL correctness
Proxy settings that might interfere

Error Messages

Error	Likely Cause
`authentication failed`	Missing or invalid API key
`model not found`	Incorrect model ID or unavailable model
`rate limit exceeded`	Too many requests, wait and retry
`connection refused`	Local server not running or wrong port
`timeout`	Request exceeded configured timeout
`chunk timeout`	Streaming response stalled

Provider Configuration Inheritance

When multiple models from the same provider share options, you can set provider-level defaults that apply to all models:

{
  "provider": {
    "anthropic": {
      "options": {
        "timeout": 300000,
        "setCacheKey": true
      }
    }
  }
}

Individual model configurations can override provider-level options.

Supported Providers​

Adding Credentials​

Authentication Methods​

Provider Configuration​

Provider Options Reference​

Provider-Specific Options​

Custom Providers​

Custom Provider Fields​

Custom Headers​

Local Model Providers​

Ollama​

LM Studio​

llama.cpp​

Atomic Chat​

Base URL Customization​

Troubleshooting​

Check Stored Credentials​

Verify Provider ID​

Testing Provider Connectivity​

Error Messages​

Provider Configuration Inheritance​