YBacked by Y Combinator

Guard. Verify. Redact.

Superagent provides purpose-trained AI models for building securely and staying compliant. Low-latency, production-ready, and simple to deploy.

10K+ GitHub Stars
Open-source (MIT)

redact catches data leaks before your customers do

Requests, responses, and tool calls are analyzed in real time, with sensitive data removed before it leaves your environment.

GET /v1/customers/cus_Qr...
{
  "id": "cus_Qr8Tx2mKLp9nF3",
  "object": "customer",
  "address": {
    "line1": "742 Evergreen Terrace",
    "city": "Springfield",
    "state": "CA",
    "postal_code": "94102",
    "country": "US"
  },
  "balance": 0,
  "created": 1704067200,
  "currency": "usd",
  "default_source": "card_1PxKm2L3nF9Tx8Y",
  "delinquent": false,
  "description": "Enterprise customer",
  "email": "sarah.chen@acmecorp.com",
  "invoice_prefix": "E8F4A2C",
  "invoice_settings": {
    "custom_fields": null,
    "default_payment_method": "pm_1PxKm2L3nF9Tx8Y",
    "footer": null
  },
  "livemode": true,
  "metadata": {
    "internal_user_id": "usr_8f3m9k2n5p7q4",
    "salesforce_id": "0018000001pZx7QAAS",
    "api_key": "sk_live_51PxKmL3nF9Tx8Y2mKLp3"
  },
  "name": "Sarah Chen",
  "phone": "+1-555-0147",
  "preferred_locales": [
    [
    ]
  ],
  "shipping": null,
  "tax_exempt": "none"
}
Monitoring
1,370 events analyzed

Monitoring...

guard stops attacks before they execute

Prompt injections, backdoors, and jailbreaks are intercepted as they happen, blocking malicious input at runtime.

verify keeps every output aligned with your truth

Model responses are continuously checked against trusted sources to ensure accuracy and compliance before delivery.

Integrate in minutes

API

Add capabilities to any system with a single HTTP request. Language-agnostic and framework-agnostic. Works with existing infrastructure without code changes.

SDKs

Native Python and TypeScript libraries for seamless integration. Embed security checks directly into your application with typed responses and async support.

CLI

Command-line tool for testing and automation. Validate prompts locally, integrate with CI/CD pipelines, or batch-process data in your workflow.

Read the docs

Frequently Asked Questions

Everything you need to know about Superagent

What is Superagent?

Superagent provides capabilities that make AI secure and compliant. We offer three purpose-trained models — Guard, Verify, and Redact — available as standalone APIs that protect your AI applications in real time with low latency.

What do Guard, Verify, and Redact do?

Guard detects and blocks unsafe inputs, prompt injections, malicious tool calls, and backdoors before they reach your models. Verify validates model outputs against your enterprise sources and policies to ensure every generation is factual and compliant. Redact removes sensitive data like PII, PHI, and secrets from text, logs, or documents automatically. All three models work with both structured and unstructured data.

How do I use Superagent?

You can integrate Superagent capabilities through our API (send any payload and receive guarded, verified, or redacted results), SDKs (Python and TypeScript libraries for embedding into workflows), CLI (command-line tool for testing and automation), or Playground (interactive web interface to explore capabilities before integration). See our documentation for detailed integration guides.

Will Superagent slow down my application?

No. Our purpose-trained models are optimized for production speed and deliver low-latency protection, so your agents and copilots stay fast while gaining enterprise-grade security.

How accurate is Guard at detecting threats?

Our guard model achieves 98% detection accuracy on internal evaluation benchmarks, outperforming GPT-5 (94.5%), Gemini 2.5 Pro (97%), Sonnet-4 (37%), and Opus 4.1 (24.5%). This means fewer missed exploits and stronger protection for production applications.

Can I use Superagent with my existing AI stack?

Yes. Superagent is language-agnostic and framework-agnostic. It works with any LLM provider (OpenAI, Anthropic, open-source models) and integrates with agent frameworks like LangGraph, Mastra, and Vercel AI SDK. Drop it into your existing infrastructure through simple API calls.

What's the difference between hosted and self-hosted deployment?

Hosted launches in seconds with zero infrastructure management and automatic scaling — ideal for fast deployment and SaaS teams. Self-hosted deploys inside your VPC for full data ownership and enterprise-grade controls — perfect for teams with strict data residency requirements.

Do you support compliance frameworks?

Yes. Superagent provides a Trust Center — a public compliance hub that shows how Guard, Verify, and Redact map to frameworks like the EU AI Act, ISO/IEC 42001, and NIST AI RMF. This transparency layer helps you pass enterprise security reviews and prove compliance to customers and auditors.

Is Superagent open-source?

Yes. Superagent is released under the MIT license with over 10k stars on GitHub. Our guard models, evaluation datasets, and benchmarks are available on HuggingFace for teams that want full transparency and control.

What use cases does Superagent support?

Superagent is built for organizations deploying AI in production. Common use cases include: runtime protection for deployed agents against prompt injections and malicious tool calls, continuous verification to ensure model outputs align with company or regulatory sources, and input/output sanitization to redact PII and sensitive data automatically before or after AI processing.

How do I get started?

Sign up at app.superagent.sh to start using Guard, Verify, and Redact in minutes. For teams evaluating Superagent, book a 30-minute call to discuss your specific security and compliance requirements.

Recent blog posts

The case for small language models

Most agents today rely on large, general-purpose models built to do everything. If your agent has a single, well-defined job, it should also have a model designed for that job. This is the case for small language models: models that handle one task, run locally, and can be retrained as your data evolves.

engineering·

Why Your AI Agent Needs More Than Content Safety

You've enabled Azure Content Safety or Llama Guard. Your AI agent still isn't secure. Here's why content filtering isn't enough when your AI takes actions.

security·

Shipped: Runtime Redaction and Command-Line Security

The past two weeks brought runtime redaction, a powerful CLI, URL whitelisting, and a developer experience that puts security directly in your workflow. Here's what shipped and why it matters for teams building with AI agents.

announcements·

Three years later: AI can (now) defend AI

In 2022, Simon Willison argued that 'adding more AI' was the wrong fix for prompt injection and related failures. He was mostly right at the time. What people tried then were brittle ideas that either overblocked or were easy to trick. This post explains what has changed since, what has not, and why builders can now use AI to meaningfully defend their agents in production.

security·
View more posts

Become secure and compliant