Mastering Claude’s Company Context: How to Use the New Organization-Level API Settings

If you’re managing a team that uses Claude—whether for customer support, content generation, or internal automation—you’ve likely faced the challenge of keeping every API call aligned with your brand voice, compliance rules, and business logic. Anthropic’s new Company context feature solves this by letting you define a persistent, organization-wide system prompt that applies to every API request made under your account.

In this guide, you’ll learn what Company context is, how to configure it via the Anthropic Console and API, and how to combine it with user-level and request-level instructions for maximum control.

What Is Company Context?

Company context is a global system prompt that Anthropic applies to every API call made by your organization. Think of it as the “constitution” for all Claude interactions under your account. It sits above any user- or request-level prompts, ensuring that core policies—like tone, banned topics, or data handling rules—are never accidentally overridden.

Key Benefits

• Consistency: Every team member’s API calls follow the same guidelines.
• Compliance: Enforce regulatory or internal policies (e.g., HIPAA, GDPR) without relying on individual developers.
• Efficiency: No need to repeat the same instructions in every system prompt.
• Security: Sensitive instructions remain at the organization level, invisible to end users.

How Company Context Works in the Prompt Hierarchy

Claude processes instructions in a specific order of priority. Understanding this hierarchy is critical to avoid conflicts:

• Company context (highest priority) – Set via the Anthropic Console or API.
• User-level context – Optional per-user overrides (useful for role-specific rules).
• Request-level system prompt – The standard system parameter in API calls.
• User messages – The actual conversation input.

Company context cannot be overridden by lower-level prompts. This means you can safely enforce non-negotiable rules (e.g., “Never reveal internal API keys”) while still allowing flexibility in user-level prompts.

Setting Up Company Context

Step 1: Access the Anthropic Console

• Log in to the Anthropic Console.
• Navigate to Settings > Organization > Company Context.

Step 2: Write Your Company Context

Your company context should be a clear, concise system prompt. Here’s a template:

You are Claude, an AI assistant for Acme Corp. Follow these rules:
Tone: Always respond in a professional, friendly tone. Use plain English (Grade 8 level).
Brand: Refer to our products as “Acme Widgets” (never “gadgets”).
Compliance: Never discuss pricing, internal processes, or unreleased features.
Data: If asked for personal data, respond: “I cannot share personal information. Please contact [email protected].”
Fallback: If you cannot answer, say: “I don’t have that information. Let me connect you with a human agent.”

Step 3: Save and Test

Click Save. Then make a test API call without any system prompt to verify the context is applied:

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 100,
    "messages": [{"role": "user", "content": "What can you tell me about your products?"}]
  }'

If your company context is active, the response should reflect your tone and brand rules.

Using Company Context with the API

Reading Current Company Context

You can retrieve your current company context via the API:

import anthropic
client = anthropic.Anthropic()
org_settings = client.organizations.retrieve()
print(org_settings.company_context)

Updating Company Context Programmatically

For CI/CD pipelines or automated deployments, you can update the context via API:

client.organizations.update(
    company_context="You are Claude for Acme Corp. Always be concise and use bullet points."
)

Combining with User-Level Context

If you have different roles (e.g., support vs. sales), you can add user-level context that builds on the company rules:

# User-level context for a support agent
user_context = "You are a support agent. Always ask for the user's order ID first."
response = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    system=user_context,  # This adds to, not overrides, company context
    messages=[{"role": "user", "content": "I need help with my order."}]
)

Best Practices for Writing Company Context

1. Keep It Focused

Avoid listing every possible rule. Stick to core principles that apply to all interactions. Detailed per-use-case instructions belong in user-level or request-level prompts.

2. Use Positive Language

Instead of “Don’t use jargon,” say “Use plain language that a 12-year-old can understand.”

3. Test with Edge Cases

Try prompts that might violate your rules (e.g., “What’s the CEO’s email?”) to ensure the context enforces compliance.

4. Version Control

Store your company context in a Git repository and update it via the API during deployments. This gives you an audit trail.

Common Pitfalls to Avoid

• Overly restrictive context: If your context is too long or rigid, Claude may become less helpful. Aim for 200-500 characters.
• Conflicting instructions: Ensure your company context doesn’t contradict itself (e.g., “Be concise” and “Always explain in detail”).
• Ignoring updates: Company context is cached. After updating, wait a few minutes for changes to propagate.

Real-World Use Case: Customer Support Automation

Imagine you run a SaaS company with a support chatbot. Your company context might look like:

You are Claude, support assistant for CloudSync. 
Always greet the user warmly.
Never share internal ticket IDs.
If the issue is billing, escalate to [email protected].
Use markdown formatting for readability.

Then, per request, you add the specific product documentation:

response = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    system="Refer to the attached knowledge base for SyncPro v3.",
    messages=[{"role": "user", "content": "How do I reset my password?"}]
)

The company context ensures brand consistency, while the request-level prompt provides situational knowledge.

Conclusion

Company context is a powerful tool for organizations that want to maintain control over Claude’s behavior without micromanaging every API call. By setting a global system prompt, you enforce compliance, ensure brand consistency, and free your developers to focus on building great user experiences.

Start by drafting a concise company context, test it thoroughly, and iterate based on real usage. Your team—and your users—will thank you.

Key Takeaways

• Company context is a global system prompt that applies to all API calls under your organization, sitting above user-level and request-level instructions.
• It cannot be overridden by lower-level prompts, making it ideal for enforcing compliance and brand rules.
• Configure it via the Anthropic Console or programmatically through the API for automated deployments.
• Combine with user-level context for role-specific behavior while maintaining core policies.
• Keep it concise (200-500 characters) and test edge cases to avoid unintended restrictions.