Runtime protection for AI agents and copilots - inspect prompts, validate tool calls, and block threats in real time.
Superagent is a secure proxy between your apps, models, and tools. It provides:
- Runtime Protection - Detects prompt injections, backdoors, and data leaks in real time
- Guarded Tooling - Validates tool calls and parameters before execution
- SuperagentLM - Reasoning-driven safety model with sub-50ms latency
- Unified Observability - Centralized policies, audits, and compliance logs
This repository contains:
- SDKs (Python & TypeScript) - Validate actions directly in your app without running the proxy
- Proxy (Node.js & Rust) - Route requests through a secure reverse proxy with runtime protection
Install and use the SDK to enforce runtime protections directly in your application:
Python:
pip install superagent-ai
TypeScript:
npm install superagent-ai
See sdk/python/README.md and sdk/typescript/README.md for usage examples and API documentation.
Run Superagent as a reverse proxy to protect all AI API traffic:
Node.js:
cd proxy/node/
npm install
npm start
Rust (High Performance):
cd proxy/rust/
cargo build --release
./target/release/ai-firewall start
Docker:
docker-compose up -d
See proxy/node/README.md and proxy/rust/README.md for configuration, CLI options, and programmatic usage.
Edit superagent.yaml
to configure models and providers:
models:
- model_name: "gpt-5"
provider: "openai"
api_base: "https://api.openai.com"
- model_name: "claude-sonnet-4-5"
provider: "anthropic"
api_base: "https://api.anthropic.com/v1"
# Optional: Send telemetry to external webhook
telemetry_webhook:
url: "https://your-webhook.com/api/telemetry"
headers:
x-api-key: "your-api-key"
├── proxy/
│ ├── node/ # Node.js proxy implementation
│ └── rust/ # Rust proxy implementation (high performance)
├── sdk/
│ ├── python/ # Python SDK
│ └── typescript/ # TypeScript SDK
├── docker/ # Docker configurations
└── README.md # This file
- Docs: docs.superagent.sh
- Models: huggingface.co/superagent-ai
See LICENSE file for details.