Getting Started with AgentWall in 5 Minutes

AgentWall integration takes minutes, not days. This guide walks you through setup, from account creation to your first protected agent run. In 5 minutes, you'll have comprehensive governance, security, and cost controls.

What You'll Get

After completing this guide, your agents will have:

• Cost controls: Automatic budgets and kill switches
• Security: DLP, prompt injection detection, and API key protection
• Observability: Complete visibility into agent operations
• Loop detection: Automatic detection and termination of infinite loops
• Audit trails: Comprehensive logging for compliance

Step 1: Create Account (30 seconds)

Visit agentwall.io and sign up. No credit card required for the free tier. You'll get:

• 10,000 free requests per month
• Full feature access
• Dashboard and analytics
• Email support

Step 2: Get Your API Key (15 seconds)

After signup, navigate to Settings → API Keys. Click "Create New Key" and copy it. This key authenticates your requests to AgentWall.

Store the key securely—treat it like a password. Add it to your environment variables:

AGENTWALL_API_KEY=aw_your_key_here

Step 3: Update Your Code (2 minutes)

For OpenAI SDK

Change your base URL to point to AgentWall. That's it—no other code changes needed.

from openai import OpenAI

client = OpenAI(
    api_key="your-openai-key",
    base_url="https://api.agentwall.io/v1",
    default_headers={
        "X-AgentWall-Key": "aw_your_key_here"
    }
)

# Use normally
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}]
)

For LangChain

Set the openai_api_base parameter:

from langchain.chat_models import ChatOpenAI

llm = ChatOpenAI(
    openai_api_key="your-openai-key",
    openai_api_base="https://api.agentwall.io/v1",
    headers={"X-AgentWall-Key": "aw_your_key_here"}
)

For Direct API Calls

Route requests through AgentWall:

curl https://api.agentwall.io/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your-openai-key" \
  -H "X-AgentWall-Key: aw_your_key_here" \
  -d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Step 4: Configure Policies (2 minutes)

Set Budgets

In the AgentWall dashboard, navigate to Policies → Budgets. Set limits:

• Per-run budget: Maximum cost for a single agent task (e.g., $5)
• Daily budget: Maximum daily spending (e.g., $100)
• Monthly budget: Total monthly limit (e.g., $1000)

AgentWall enforces these budgets automatically. Requests are blocked when limits are reached.

Enable Security

Navigate to Policies → Security. Enable protections:

• DLP: Detect and redact PII, API keys, and sensitive data
• Prompt injection detection: Block malicious prompts
• Output filtering: Prevent sensitive data in responses

Configure Loop Detection

Go to Policies → Loop Detection. Set thresholds:

• Max steps: Maximum steps per run (e.g., 50)
• Repetition threshold: How many repeated actions trigger alerts (e.g., 5)
• Auto-kill: Automatically terminate loops (recommended)

Step 5: Test Your Integration (30 seconds)

Run a test request through your updated code. Check the AgentWall dashboard—you should see:

• The request logged
• Cost calculated
• Security checks passed
• Run tracked

If you see your request in the dashboard, integration is successful!

Advanced Configuration

Run Identification

For better tracking, include run IDs in your requests:

response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    extra_headers={
        "X-AgentWall-Run-ID": "unique-run-id-123"
    }
)

Run IDs enable tracking complete agent tasks across multiple requests.

Metadata Tagging

Add metadata for better organization:

extra_headers={
    "X-AgentWall-Team": "engineering",
    "X-AgentWall-Project": "customer-support",
    "X-AgentWall-Environment": "production"
}

Metadata enables cost allocation, filtering, and analysis.

Custom Policies

Create agent-specific policies in the dashboard. Different agents can have different budgets, security rules, and loop detection settings.

Monitoring and Alerts

Dashboard

The AgentWall dashboard shows real-time metrics:

• Current spending rate
• Active runs
• Recent alerts
• Cost trends
• Security events

Alerts

Configure alert channels in Settings → Alerts:

• Email: Get notified of budget violations and security events
• Slack: Real-time alerts in your team channel
• Webhook: Integrate with PagerDuty or custom systems

Common Issues

Authentication Errors

Problem: "Invalid AgentWall API key"

Solution: Verify your AgentWall key is correct and included in the X-AgentWall-Key header.

Requests Not Appearing

Problem: Requests work but don't show in dashboard

Solution: Check that base_url points to AgentWall. Verify the AgentWall key is included.

Budget Blocks

Problem: Requests blocked with "Budget exceeded"

Solution: Increase budgets in dashboard or wait for budget reset (daily/monthly).

Next Steps

Explore Features

Now that basic integration is complete, explore advanced features:

• Execution replay for debugging
• Custom security rules
• Team management and access controls
• Cost allocation by team or project
• API for programmatic access

Optimize Policies

Review your initial policies after a few days. Adjust budgets based on actual usage. Tune security rules to reduce false positives.

Add More Agents

Integrate all your agents with AgentWall. Consistent governance across all agents simplifies management and ensures comprehensive protection.

Getting Help

Documentation

Visit docs.agentwall.io for comprehensive documentation: API reference, integration guides, and best practices.

Support

Contact support@agentwall.io for help. Free tier includes email support. Paid plans include priority support and Slack access.

Community

Join the AgentWall community on Discord. Share experiences, ask questions, and learn from other users.

Conclusion

You're now running protected, governed AI agents. AgentWall provides the security, cost controls, and observability needed for production AI operations.

Welcome to the AgentWall community. Guard the Agent, Save the Budget!