Securing MCP-Based LLM Apps
Request → Pre-call Validation → MCP Tool → Post-call Filtering → Safe Output.
- Developed security controls for an MCP/FastMCP AI assistant.
- Implemented pre-call and post-call validation to detect unsafe inputs and filter risky outputs.
- Tested for prompt injection, XSS, SQL injection, secret leakage and suspicious API behaviour.
- Contributed to testing, documentation and prototype development.