5-Minute Quickstart

Secure your AI agent with one install command and fewer than 5 lines of code. No complex infrastructure or proprietary YAML required.

1

Sign up & Get API key

Create an account and copy your SupraWall API key from the dashboard.

Your API key is ready

Prefix: ag_live_...

Go to dashboard
2

Install the SDK

Native packages available for all major environments.

BASH
npm install suprawall
3

Initialize Client

Set your security posture to 'Deny-by-default' for zero-trust protection.

TYPESCRIPT
import { Client } from "suprawall";
    
const client = new Client({
  apiKey: process.env.SUPRAWALL_API_KEY,
  defaultPolicy: "DENY" // Every action is blocked unless explicitly allowed
});
4

Wrap your agent

Connect the security layer to your agent executor. Every tool call is now intercepted.

TYPESCRIPT
import { secure_agent } from "suprawall";
    
// Wrap your existing agent (LangChain, Vercel AI, etc.)
const secured = secure_agent(myAgent, { client });
    
// Execution is now gated by your policies
await secured.invoke({ task: "Delete production database" });
5

Define Policy & Test

Add a simple policy in the dashboard to block specific tools and verify they are gated.

Blocked Tool Example

Try calling a tool that should be restricted, like `db.drop_table`:

PYTHON
try:
    response = secured.invoke({"task": "drop the users table"})
except Exception as e:
    print(f"Blocked by SupraWall: {e}")

Step 6: Success!

Your agent is now bulletproof. Every blocked attempt generates a real-time audit log with a unique trace ID.

[2026-03-23 16:31:05] DENY tool:db.drop_table

Policy: restrict_db_operations

Trace ID: sup_7x2v1q9...

FAQ

What's the performance impact?

Policy evaluation is ultra-high performance, typically under 10ms. For agentic workflows which takes seconds to execute LLM calls, the impact is effectively zero.

Is it really database-agnostic?

Yes. Use our hosted Firebase cloud or bring your own Postgres, MySQL, or MongoDB via the Adapters architecture.

Do I need to change my prompt engineering?

No. SupraWall lives in the execution layer. Your prompts remain yours — we only govern the tool outputs and inputs.