Skip to main content
Claude is Anthropic’s family of AI models, known for their advanced reasoning capabilities and extended thinking modes. Avante.nvim supports both API key authentication and Claude Pro/Max subscription authentication.

​
Quick Start

​
API Key Authentication

1

Get your API key

Sign up at Anthropic Console and create an API key.
2

Set environment variable

Add to your shell configuration:
# Scoped (recommended)
export AVANTE_ANTHROPIC_API_KEY=your-api-key

# Or global
export ANTHROPIC_API_KEY=your-api-key
3

Configure provider

{
  "yetone/avante.nvim",
  opts = {
    provider = "claude",
  },
}

​
Claude Pro/Max Subscription

If you have a Claude Pro or Max subscription, you can authenticate using OAuth:
1

Enable OAuth authentication

{
  "yetone/avante.nvim",
  opts = {
    provider = "claude",
    providers = {
      claude = {
        auth_type = "max", -- Use "max" for subscription auth
      },
    },
  },
}
2

Authenticate in browser

When you open Neovim, the authentication process will start in your browser.
3

Enter authorization code

Copy the code shown in your browser and paste it into the Neovim prompt.
If you previously had a different provider selected, run :AvanteSwitchProvider claude to initiate authentication.

​
Configuration

​
Basic Configuration

providers = {
  claude = {
    endpoint = "https://api.anthropic.com",
    model = "claude-sonnet-4-5-20250929",
    timeout = 30000,
    context_window = 200000,
    extra_request_body = {
      temperature = 0.75,
      max_tokens = 64000,
    },
  },
}

​
Available Models

providers = {
  claude = {
    model = "claude-sonnet-4-5-20250929",
    extra_request_body = {
      max_tokens = 64000,
    },
  },
}

​
Authentication Types

providers = {
  claude = {
    auth_type = "api",
  },
}

​
Environment Variables

Claude uses the following environment variables:
VariableScoped VersionPurpose
ANTHROPIC_API_KEYAVANTE_ANTHROPIC_API_KEYAPI authentication
When using auth_type = "max", API keys are not required as OAuth is used instead.

​
Advanced Features

​
Prompt Caching

Claude supports prompt caching to reduce costs and latency for repeated prompts: 
providers = {
  claude = {
    -- Prompt caching is enabled by default
    support_prompt_caching = true,
  },
}
Avante automatically applies caching to:
  • System prompts
  • Tool definitions
  • The last text message in history

​
Extended Thinking

Claude models support extended thinking for complex reasoning tasks. This is automatically enabled when the model returns thinking content. 
-- Thinking is displayed with <think> tags in the output
-- Example:
-- <think>
-- Let me analyze this code structure...
-- </think>
-- 
-- Here's the solution:

​
Custom Endpoint

For self-hosted or proxy setups: 
providers = {
  claude = {
    endpoint = "https://your-custom-endpoint.com",
  },
}

​
Proxy Configuration

providers = {
  claude = {
    proxy = "http://proxy.example.com:8080",
    allow_insecure = false,
  },
}

​
Rate Limiting

Claude automatically handles rate limiting by:
  1. Detecting rate limit headers
  2. Calculating sleep time based on reset timestamps
  3. Pausing requests until limits reset
You can monitor rate limits in the response headers:
  • anthropic-ratelimit-tokens-remaining
  • anthropic-ratelimit-tokens-reset
  • anthropic-ratelimit-requests-reset

​
Troubleshooting

​
Authentication Failed

Ensure your API key is set in your environment:
echo $ANTHROPIC_API_KEY
# or
echo $AVANTE_ANTHROPIC_API_KEY
Restart Neovim after setting the variable.
If the OAuth code doesn’t work:
  1. Ensure you copied the entire code from the browser
  2. Try the authentication flow again
  3. Check that auth_type = "max" is set correctly
OAuth tokens are automatically refreshed. If you see expiration errors:
  1. Wait a moment for auto-refresh
  2. Restart Neovim to trigger re-authentication

​
Quota Exceeded

If you see quota errors: 
insufficient_quota: You don't have any credits or have exceeded your quota
Check your usage at Anthropic Console and add credits or upgrade your plan.

​
Temperature Errors

Claude requires temperature between 0 and 1: 
extra_request_body = {
  temperature = 0.75, -- Must be 0.0 to 1.0
}

​
Best Practices

Model Selection

  • Sonnet 4.5: Best balance of speed and capability
  • Opus 4: Maximum capability for complex tasks
  • Haiku 3.5: Fastest, best for simple tasks

Token Management

  • Use prompt caching for repeated contexts
  • Set appropriate max_tokens limits
  • Monitor usage in Anthropic Console

Temperature Settings

  • 0.0-0.3: Focused, deterministic responses
  • 0.4-0.7: Balanced creativity and consistency
  • 0.8-1.0: More creative and varied responses

OAuth vs API Key

  • OAuth: For Claude Pro/Max subscribers
  • API Key: For pay-per-use developers
  • Both work with same models

​
Example Configurations

{
  provider = "claude",
  providers = {
    claude = {
      model = "claude-sonnet-4-5-20250929",
      timeout = 30000,
      extra_request_body = {
        temperature = 0.7,
        max_tokens = 32000,
      },
    },
  },
}

Build docs developers (and LLMs) love

Get started for freeTalk to us