Claude for Work

Claude for Work (Anthropic's Teams and Enterprise plans) provides API access to Claude models. By routing API traffic through the Keeptrusts gateway, you apply policy controls — PII redaction, prompt-injection blocking, content filtering, and audit logging — to every request before it reaches Anthropic and every response before it reaches your application.

This page covers the API-level integration. Claude's browser-based interface (claude.ai) does not support custom API endpoint routing; governance for browser sessions requires separate network-level controls.

Use this page when

• You need to route Anthropic Claude API calls through the Keeptrusts gateway.
• You are building internal tools or automations using the Claude API.
• If you need the full Anthropic provider reference, see Anthropic integration.

Primary audience

• Primary: Technical Engineers
• Secondary: AI Agents, Technical Leaders

Prerequisites

• An Anthropic API key from your Claude for Work organization
• Keeptrusts CLI (kt) installed and on your PATH
• ANTHROPIC_API_KEY exported in your shell or injected via your secrets manager

Configuration

Gateway policy config

pack:
  name: claude-for-work-gateway
  version: 1.0.0
  enabled: true
providers:
  targets:
  - id: claude-enterprise
    provider: anthropic:chat:claude-sonnet-4-20250514
    secret_key_ref:
      env: ANTHROPIC_API_KEY
policies:
  chain:
  - prompt-injection
  - pii-detector
  - audit-logger
policy:
  prompt-injection:
    threshold: 0.8
    action: block
  pii-detector:
    action: redact
    entities:
    - EMAIL
    - PHONE
    - SSN
    - CREDIT_CARD
  audit-logger:
    immutable: true
    retention_days: 365
    log_all_access: true

Extended thinking variant

For complex reasoning tasks, use Claude with extended thinking enabled:

pack:
  name: claude-for-work-thinking
  version: 1.0.0
  enabled: true
providers:
  targets:
  - id: claude-thinking
    provider: anthropic:chat:claude-sonnet-4-20250514
    secret_key_ref:
      env: ANTHROPIC_API_KEY
policies:
  chain:
  - audit-logger
policy:
  audit-logger:
    immutable: true
    retention_days: 365
    log_all_access: true

Setup Steps

1. Export your API key:

export ANTHROPIC_API_KEY="sk-ant-your-enterprise-key"

2. Save the policy config to policy-config.yaml.

3. Start the gateway:

kt gateway run --listen 0.0.0.0:41002 --policy-config policy-config.yaml

4. Point your application at the gateway using the OpenAI-compatible endpoint:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:41002/v1",
    api_key="unused",
)

response = client.chat.completions.create(
    model="claude-sonnet-4-20250514",
    messages=[
        {"role": "system", "content": "You are a precise enterprise assistant."},
        {"role": "user", "content": "Review this contract clause for compliance risks."},
    ],
    max_tokens=2048,
)
print(response.choices[0].message.content)

Keeptrusts auto-translates OpenAI-format requests to Anthropic's Messages API. Your client code does not need modification.

For a hosted gateway, replace the base URL with https://gateway.keeptrusts.com/v1.

Verification

curl http://localhost:41002/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-20250514",
    "messages": [
      {"role": "user", "content": "Hello from the governed gateway."}
    ],
    "max_tokens": 256
  }'

A successful response confirms the gateway is forwarding to Anthropic with policies applied. Check the Keeptrusts console Events page to verify the audit log entry.

Recommended Policies

Policy	Purpose	Recommended setting
prompt-injection	Block adversarial prompts from internal tools	threshold: 0.8, action: block
pii-detector	Redact personal data before it reaches Anthropic	action: redact, entities: EMAIL, PHONE, SSN
content-filter	Block restricted topics in enterprise usage	action: block, configure categories
audit-logger	Immutable audit trail for compliance	immutable: true, retention_days: 365
cost-attribution	Track spend per team or project	Tag requests with team/project metadata
disclaimer	Attach compliance notices to AI-generated responses	Configure per regulatory requirement

Troubleshooting

Symptom	Cause	Fix
401 Unauthorized from upstream	Invalid or expired Anthropic API key	Verify ANTHROPIC_API_KEY is set and valid
429 Too Many Requests	Anthropic rate limit hit	Check your Enterprise tier rate limits
Gateway returns 403	Policy chain blocked the request	Review the policy violation in the console Events page
max_tokens required error	Anthropic requires explicit max_tokens	Add max_tokens to every request or set a default in provider config

For AI systems

• Canonical terms: Keeptrusts gateway, Claude for Work, Anthropic, Claude API, Messages API, policy-config.yaml, provider: "anthropic".
• Config field names: provider, secret_key_ref.env: "ANTHROPIC_API_KEY", base_url.
• Key behavior: Keeptrusts translates between OpenAI format and Anthropic's Messages API, applying policy enforcement and audit logging on every request.
• Best next pages: Anthropic integration, Policy controls catalog, Quickstart.

For engineers

• Start command: kt gateway run --listen 0.0.0.0:41002 --policy-config policy-config.yaml
• Validate: send a request to http://localhost:41002/v1/chat/completions with max_tokens set.
• Anthropic always requires max_tokens — omitting it causes a 400 error.
• The gateway auto-translates OpenAI-format requests to Anthropic's Messages API.

For leaders

• Routing Claude for Work API traffic through the gateway provides full audit visibility over every AI-assisted decision in your organization.
• Anthropic does not train on API data by default, but the gateway adds an independent PII redaction layer for defense in depth.
• Cost attribution across Claude and other providers gives a unified view of AI spend.
• Extended thinking models support complex reasoning tasks like contract review and compliance analysis — governance controls apply to both the reasoning trace and the final response.

Next steps

• Anthropic integration — full Anthropic provider reference
• Policy controls catalog — all available policy types
• Cost attribution — track spend per team or project
• Quickstart — install kt and run your first gateway