superagent-ai/superagent

Project overview

A library and CLI that embeds runtime safety into AI applications to block prompt injections, redact sensitive information, and scan repositories for threats. Developers can integrate it via TypeScript or Python SDKs, run open-weight models on their own infrastructure, or use it as an MCP server with Claude tools.

Project type

MCP · AI Agent · Infrastructure

Use cases

Automation

Deployment

Refer to project documentation

License

MIT

Best for

• Developers and AI engineers building applications who need to integrate runtime safety checks via an SDK or CLI.
• Organizations that require data boundary control and need to run security models on their own infrastructure using open-weight options.

Key capabilities

• Detects and blocks prompt injections, malicious instructions, and unsafe tool calls during runtime.
• Removes personally identifiable information, protected health information, and secrets from text automatically.
• Analyzes code repositories for AI agent-targeted attacks, such as repository poisoning and embedded malicious instructions.
• Runs red team scenarios against production agents to identify vulnerabilities.
• Provides a command-line tool for testing and automation of security workflows.
• Offers an MCP server to integrate functionality with Claude Code and Claude Desktop.
• Provides open-weight models to run Guard on own infrastructure with no data leaving the environment.

Limitations and risks

• The capability to run red team scenarios against production agents is not yet available and is coming soon.

Getting started

• Sign up at superagent.sh to obtain an API key, install the package via npm or uv, and set the SUPERAGENT_API_KEY environment variable.

Evidence and sources

• GitHub project description: Superagent protects your AI applications against prompt injections, data leaks, and harmful outputs. Embed safety directly into your app and prove compliance to your customers.
• README: An open-source SDK for AI agent safety. Block prompt injections, redact PII and secrets, scan repositories for threats, and run red team scenarios against your agent.
• README: <strong>Make your AI apps safe.</strong>
• README: Detect and block prompt injections, malicious instructions, and unsafe tool calls at runtime.
• README: Remove PII, PHI, and secrets from text automatically.

Reviewed at: 2026-07-15T15:34:08Z