Introduction

When deploying Claude API across a team or organization, one of the biggest challenges is maintaining consistent behavior. Without a shared context, each developer might craft prompts that drift from your brand voice, ignore compliance rules, or miss critical business logic. Anthropic’s Company changelog feature addresses this by allowing you to define persistent context that applies to all API requests under your organization.

In this guide, you’ll learn what the Company changelog is, how to configure it, and how to use it effectively in your Claude API workflows. We’ll walk through real-world examples in Python and TypeScript, and share best practices for keeping your organizational context up to date.

What Is the Company Changelog?

The Company changelog is a feature within the Anthropic Console that lets you set a global system prompt or context for your entire organization. Think of it as a shared “instruction manual” that Claude reads before processing any API call from your team. This is especially useful for:

• Brand voice enforcement – Ensure all responses match your tone guidelines.
• Compliance and policy – Inject legal disclaimers or content restrictions.
• Domain-specific knowledge – Provide background on your products, services, or industry.
• Multi-user consistency – Eliminate prompt drift across different developers.

Setting Up Your Company Context

Step 1: Access the Anthropic Console

Navigate to the Anthropic Console and log in with your organization’s admin account. Go to the Settings or Company section (the exact label may vary as the UI evolves).

Step 2: Edit the Company Changelog

You’ll see a text area labeled something like “Company Context” or “Organization Instructions.” This is your changelog. Write your global instructions in plain text or Markdown. For example:

You are Claude, an AI assistant for Acme Corp. Always:
Address customers politely and use a friendly, professional tone.
Never share internal financial data or unreleased product plans.
If asked about pricing, refer to the public pricing page at https://acme.com/pricing.
When writing code, prefer Python and include type hints.

Step 3: Save and Test

Click Save. Now every API call made with your organization’s API key will include this context automatically. Test it with a simple request:

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-3-opus-20240229",
    "max_tokens": 100,
    "messages": [{"role": "user", "content": "What is your pricing?"}]
  }'

The response should reflect your company context, e.g., “For pricing details, please visit our public pricing page at https://acme.com/pricing.”

Using Company Context in Your Code

Once your company context is set, you don’t need to repeat it in every API call. However, you can still override or extend it per request using the system parameter. Here’s how to work with it in Python and TypeScript.

Python Example

import anthropic
client = anthropic.Anthropic(api_key="your-api-key")
Company context is automatically included
message = client.messages.create(
    model="claude-3-sonnet-20240229",
    max_tokens=200,
    messages=[
        {"role": "user", "content": "Write a welcome email for a new customer."}
    ]
)
print(message.content[0].text)

If you need to add per-request instructions, use the system parameter:

message = client.messages.create(
    model="claude-3-sonnet-20240229",
    system="IMPORTANT: This email is for a VIP customer who just spent over $10,000.",
    max_tokens=200,
    messages=[
        {"role": "user", "content": "Write a welcome email for a new customer."}
    ]
)

The system prompt merges with your company context (company context is applied first, then the per-request system prompt).

TypeScript Example

import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY,
});
async function generateEmail() {
  const message = await client.messages.create({
    model: 'claude-3-sonnet-20240229',
    max_tokens: 200,
    messages: [
      { role: 'user', content: 'Write a welcome email for a new customer.' }
    ],
  });
console.log(message.content[0].text);
}
generateEmail();

Again, company context is automatically injected. To add per-request instructions:

const message = await client.messages.create({
  model: 'claude-3-sonnet-20240229',
  system: 'IMPORTANT: This email is for a VIP customer who just spent over $10,000.',
  max_tokens: 200,
  messages: [
    { role: 'user', content: 'Write a welcome email for a new customer.' }
  ],
});

Best Practices for Managing Company Context

1. Keep It Concise

Your company context should be a single, focused block of instructions. Avoid long-winded explanations. Claude has a limited context window, and every token counts. Aim for 200–500 words.

2. Use Version Control

Treat your company context like code. Maintain a changelog of edits in your own repository (e.g., company-context-v1.md, v2.md). This helps you roll back if a change causes unexpected behavior.

3. Test Changes in a Staging Environment

Before updating the live company context, test your new instructions in a separate API key or a development organization. This prevents breaking production workflows.

4. Combine with Per-Request Overrides

Use the company context for stable, long-term rules (brand voice, compliance). Use per-request system prompts for temporary or task-specific instructions (e.g., “Today’s promotion is 20% off.”).

5. Monitor and Iterate

Review the output of your API calls regularly. If Claude starts ignoring your company context, it might be too verbose or contradictory. Simplify and clarify.

Common Pitfalls

• Conflicting instructions – If your company context says “be concise” but a per-request prompt says “provide detailed explanations,” Claude may struggle. Keep instructions aligned.
• Outdated context – If your company changes its pricing or policies, update the changelog immediately. Otherwise, Claude will give incorrect information.
• Over-reliance on context – Don’t expect the company context to fix poorly designed prompts. It’s a supplement, not a substitute for good prompt engineering.

Conclusion

Anthropic’s Company changelog is a powerful tool for teams that want to deploy Claude API at scale with consistent behavior. By setting a global context, you reduce prompt duplication, enforce policies, and ensure every response aligns with your organization’s standards. Combine it with per-request overrides for maximum flexibility, and keep your context updated as your business evolves.

Key Takeaways

• The Company changelog lets you set persistent, organization-wide instructions for all Claude API calls.
• You can override or extend the company context per request using the system parameter in the Messages API.
• Keep your company context concise (200–500 words) and version-controlled for easy rollback.
• Always test context changes in a staging environment before applying them to production.
• Combine stable company rules with dynamic per-request prompts for the best results.