How to Use VeloFill with Claude (Anthropic)

VeloFill connects to any OpenAI-compatible LLM provider, giving you the freedom to choose the best AI for your form-filling workflows. Anthropic’s Claude models are among the most capable options available, with strong reasoning, excellent instruction-following, and robust handling of complex forms. The challenge: Anthropic’s native API uses a different format than OpenAI’s.

This guide shows you three reliable paths to use Claude with VeloFill:

• OpenRouter (recommended for individuals): A unified API that provides OpenAI-compatible access to Claude and hundreds of other models with no verification delays.
• Direct Anthropic API: Connect directly to Anthropic’s servers using their OpenAI-compatible endpoint. Best for users who want the latest features and transparent pricing.
• LiteLLM proxy (recommended for teams): A self-hosted gateway that translates OpenAI-format requests to Anthropic’s native API.

All three approaches let you leverage Claude’s capabilities while keeping VeloFill’s architecture simple and your data private.

UI note (current extension versions): In the Connections form, VeloFill includes built-in provider presets (OpenAI, Anthropic Claude, Cohere, Ollama, Groq, Together AI, Mistral AI, OpenRouter, Z.AI) plus Custom Provider. Use the matching preset when available, and use Custom Provider for self-hosted or unsupported endpoints.

Why Choose Claude for Form Filling

Claude models excel at the specific tasks VeloFill performs: understanding form structure, mapping knowledge base content to fields, and generating accurate, context-aware responses.

Strong instruction-following: Claude consistently adheres to system prompts and custom instructions in your Knowledge Base. This matters when you need specific formatting, tone, or compliance requirements in filled forms.

Extended context handling: Claude Sonnet 4.5 and Claude Opus 4.6 support up to 1 million tokens of context, making them suitable for large knowledge bases and complex, multi-page forms.

Reasoning capabilities: Claude’s reasoning abilities help it interpret ambiguous field labels, understand form intent, and make sensible decisions when your knowledge base doesn’t contain an exact match.

Safety and alignment: Anthropic’s focus on AI safety means Claude tends to avoid generating inappropriate content or making up information, reducing the risk of embarrassing form submissions.

Choosing the Right Claude Model

Anthropic offers several Claude model families, each optimized for different use cases. Here’s how they map to form-filling scenarios:

For most VeloFill users: Start with Claude Sonnet 4.5. It offers the best balance of capability, 1M token context, and cost for form-filling tasks.

For critical submissions: Claude Opus 4.6 delivers maximum reasoning capability for complex applications, legal forms, or situations where accuracy is paramount. It’s Anthropic’s flagship model released in February 2026.

For high-volume workflows: Claude Haiku 4.5 provides near-frontier intelligence at lower cost and latency, suitable for simple contact forms, logins, and routine submissions.

Choosing Your Connection Path

Recommended: Start with OpenRouter if you want the fastest setup with no verification. Choose Direct Anthropic API if you have an Anthropic account and want the latest features. Use LiteLLM if you’re setting up for a team.

Path A: Connect Claude via OpenRouter (Recommended for Individuals)

OpenRouter provides a unified, OpenAI-compatible API that routes requests to Claude (and other models) behind the scenes. It’s the fastest way to get started with Claude in VeloFill.

Why OpenRouter

• OpenAI-compatible endpoint: Works directly with VeloFill’s built-in OpenRouter preset.
• No account verification delays: Unlike direct Anthropic API access, OpenRouter doesn’t require identity verification for most users.
• Automatic fallbacks: If one provider is unavailable, OpenRouter can route to alternatives.
• Usage-based pricing: Pay only for what you use, with transparent per-token rates.

Prerequisites

• An OpenRouter account (free to create at openrouter.ai)
• The VeloFill extension installed in Chrome, Firefox, or Edge

Step 1: Create an OpenRouter API Key

1. Sign in to your OpenRouter account.
2. Navigate to Keys in your dashboard.
3. Click Create Key.
4. Give your key a descriptive name like VeloFill Extension.
5. Optionally set a credit limit to control spending.
6. Copy the generated key immediately. OpenRouter won’t show it again.

⚠️ Security note: Treat your API key like a password. Store it in a password manager and never commit it to version control.

Step 2: Configure VeloFill for OpenRouter

1. Open VeloFill’s Options page (right-click the extension icon and select “Options”).
2. Navigate to the Connections tab.
3. Click Add Connection or select an existing connection to edit.
4. Configure the connection:
   • Provider: Select OpenRouter from the dropdown.
   • Connection Name: Enter a descriptive name like Claude via OpenRouter.
   • Endpoint URL: Enter https://openrouter.ai/api/v1 (see the OpenRouter quickstart guide for details).
   • API Key: Paste your OpenRouter API key.
   • Model ID: Enter the Claude model you want to use:
     • anthropic/claude-opus-4.6 for the latest flagship model
     • anthropic/claude-sonnet-4.5 for balanced performance
     • anthropic/claude-haiku-4.5 for speed and efficiency
5. Click Save.

VeloFill auto-fills the OpenRouter endpoint for this preset. If you need a non-standard gateway URL, choose Custom Provider and enter your endpoint manually.

Step 3: Test Your Connection

Before relying on Claude for important forms, verify the connection works:

1. Navigate to a simple web form (a contact form or newsletter signup works well).
2. Click the VeloFill icon and select the form.
3. Trigger the autofill and review the results.
4. If fields populate correctly, your connection is working.

Path B: Connect Claude via LiteLLM Proxy (Recommended for Teams)

LiteLLM is an open-source proxy server that translates OpenAI-format requests to Anthropic’s native API. It’s ideal for teams that want:

• Self-hosted control: Run the proxy on your own infrastructure.
• Centralized key management: Share one Anthropic API key across multiple users.
• Usage tracking: Monitor and log all requests for auditing.
• Multi-provider routing: Configure fallbacks between Claude and other models.

Prerequisites

• An Anthropic API key (from console.anthropic.com)
• Python 3.8+ installed
• Basic familiarity with command-line tools

Step 1: Install LiteLLM

pip install litellm

Step 2: Configure the Proxy

Create a configuration file named litellm_config.yaml:

model_list:
  - model_name: claude-opus
    litellm_params:
      model: anthropic/claude-opus-4-20250514
      api_key: "os.environ/ANTHROPIC_API_KEY"
  - model_name: claude-sonnet
    litellm_params:
      model: anthropic/claude-sonnet-4-20250514
      api_key: "os.environ/ANTHROPIC_API_KEY"
  - model_name: claude-haiku
    litellm_params:
      model: anthropic/claude-3-5-haiku-20241022
      api_key: "os.environ/ANTHROPIC_API_KEY"

This configuration creates three model endpoints: claude-opus for critical forms, claude-sonnet for everyday use, and claude-haiku for fast, simple fills. See the LiteLLM Anthropic documentation for more configuration options.

Note: LiteLLM uses Anthropic’s native model IDs (like claude-opus-4-20250514) rather than the OpenRouter-style IDs. These are stable release identifiers that map to the current model versions.

Step 3: Start the Proxy Server

Set your Anthropic API key as an environment variable, then start the proxy:

export ANTHROPIC_API_KEY="your-anthropic-api-key"
litellm --config litellm_config.yaml

The proxy will start on http://localhost:4000 by default.

Step 4: Configure VeloFill for Your LiteLLM Proxy

1. Open VeloFill’s Options page.
2. Navigate to the Connections tab.
3. Add a new connection with these settings:
   • Provider: Select Custom Provider.
   • Connection Name: Enter Claude via LiteLLM.
   • Endpoint URL: Enter http://localhost:4000 (or your server’s URL).
   • API Key: Enter any non-empty value (LiteLLM handles Anthropic authentication).
   • Model ID: Enter claude-opus, claude-sonnet, or claude-haiku (matching your config).
4. Click Save.

Step 5: Test Your Local Proxy

1. Ensure your LiteLLM proxy is running.
2. Navigate to a test form.
3. Activate VeloFill and select your LiteLLM connection.
4. Trigger the fill and verify results.

For production deployments, see our guide on routing VeloFill through LiteLLM for advanced configuration including authentication, logging, and multi-user setup.

Path C: Connect Directly to Anthropic API

Anthropic provides an OpenAI SDK compatibility layer that lets you connect directly to Claude without any middleware. This is the simplest path if you already have an Anthropic API key.

Why Direct Connection

• Access to latest features: Get new Anthropic features before they reach third-party platforms.
• Transparent pricing: Pay Anthropic’s rates directly with no markup.
• Simpler setup: No middleware or proxy required.
• Direct vendor access: Connect straight to Anthropic infrastructure without intermediate routing.

Compatibility note: Anthropic positions OpenAI SDK compatibility primarily for testing and comparison workflows. For full native Claude API features (such as prompt caching and complete feature parity), use Anthropic’s native API format where possible.

Prerequisites

• An Anthropic API key (from console.anthropic.com)
• The VeloFill extension installed in Chrome, Firefox, or Edge

Step 1: Get Your Anthropic API Key

1. Sign in to the Anthropic Console.
2. Navigate to API Keys in your account settings.
3. Click Create Key.
4. Give your key a descriptive name like VeloFill Extension.
5. Copy the generated key immediately. Anthropic only shows it once.

⚠️ Security note: Anthropic API keys start with sk-ant-. Treat this key like a password and never commit it to version control.

Step 2: Configure VeloFill for Direct Anthropic Connection

1. Open VeloFill’s Options page (right-click the extension icon and select “Options”).
2. Navigate to the Connections tab.
3. Click Add Connection or select an existing connection to edit.
4. Configure the connection:
   • Provider: Select Anthropic Claude from the dropdown.
   • Connection Name: Enter a descriptive name like Claude Direct.
   • Endpoint URL: Enter https://api.anthropic.com/v1 (see the Anthropic OpenAI compatibility docs for details).
   • API Key: Paste your Anthropic API key.
   • Model ID: Enter the Claude model you want to use:
     • claude-opus-4-6 for the latest flagship model
     • claude-sonnet-4-5 for balanced performance
     • claude-haiku-4-5 for speed and efficiency
5. Click Save.

VeloFill auto-fills Anthropic’s endpoint for this preset. If you need to route through a custom gateway, choose Custom Provider and enter your endpoint manually.

Step 3: Test Your Connection

1. Navigate to a simple web form.
2. Click the VeloFill icon and select the form.
3. Trigger the autofill and review the results.
4. If fields populate correctly, your connection is working.

Troubleshooting Common Issues

401 Unauthorized

Cause: Invalid or missing API key.

Solutions:

• Verify your API key is correctly copied (no extra spaces or line breaks).
• For OpenRouter, confirm the key is active in your dashboard.
• For Direct Anthropic API, ensure you’re using the full key starting with sk-ant-.
• For LiteLLM, ensure the ANTHROPIC_API_KEY environment variable is set.

404 Model Not Found

Cause: Incorrect Model ID.

Solutions:

• For OpenRouter, use the full model ID: anthropic/claude-sonnet-4.5 (not just claude-sonnet-4.5).
• For Direct Anthropic API, use: claude-opus-4-6, claude-sonnet-4-5, or claude-haiku-4-5.
• For LiteLLM, use the model_name you defined in your config file.
• Check that the model is available in your region and account tier.

Connection Timeout or Failed to Fetch

Cause: Network or endpoint configuration issue.

Solutions:

• Verify the Endpoint URL is correct (or that the correct preset is selected):
  • OpenRouter: https://openrouter.ai/api/v1
  • Anthropic Direct: https://api.anthropic.com/v1
  • LiteLLM: http://localhost:4000 (or your server URL)
• Check for firewall or VPN interference.
• For LiteLLM, confirm the proxy server is running.

Poor Fill Quality or Missing Data

Cause: Model limitations or knowledge base issues.

Solutions:

• Upgrade to a more capable model (Haiku → Sonnet → Opus).
• Review and expand your Knowledge Base entries.
• Add specific instructions to your knowledge base for complex fields.
• Check the troubleshooting guide for diagnostic steps.

Security and Privacy Considerations

When using Claude with VeloFill, understand the data flow:

Your Knowledge Base: Stored locally on your device with optional encryption. Never uploaded to VeloFill or third-party servers.

Form Content: When you trigger a fill, VeloFill sends:

• The form’s field structure and labels
• Relevant excerpts from your knowledge base
• Your custom prompt instructions

This data goes to your configured AI endpoint (OpenRouter, LiteLLM, or directly to Anthropic).

Anthropic’s Data Practices: Anthropic retains API data for 30 days for abuse monitoring, then deletes it. They do not use API data to train models. See their privacy policy for details.

OpenRouter’s Data Practices: OpenRouter logs basic request metadata but does not log prompts or completions by default. See their FAQ for details.

LiteLLM: When self-hosted, you control all logging and data retention.

Keep Learning

With Claude connected to VeloFill, you have access to some of the most capable AI models for form understanding and automation.

• Learn more about configuring connections
• Build an effective Knowledge Base
• Compare with other providers like Mistral AI
• Run VeloFill with local models via Ollama