Permission-Aware RAG: Making Your Internal AI Assistant Respect Document Access
How to make retrieval respect who can see what: ACL metadata, Entra ID and Cognito identity, Azure AI Search and Bedrock filters, Salesforce, audits and tests.
Writing Topic
Model Context Protocol, retrieval-augmented generation, agent security and the practical side of putting LLMs to work.
9 articles
3 articles
How to make retrieval respect who can see what: ACL metadata, Entra ID and Cognito identity, Azure AI Search and Bedrock filters, Salesforce, audits and tests.
Prompt injection, tool poisoning and MCP risks explained from an IT security angle, with practical controls to keep AI agents safe.
What agentic RAG adds to basic retrieval, when you actually need it, and lessons from building an internal company knowledge assistant.
6 articles
How MCP servers work in 2026: tools and resources, stdio vs Streamable HTTP, OAuth-based auth, the stateless spec, and where the protocol still falls short.
Add typo-tolerant search to AI apps and websites with Fuse.js, with working code, tuning tips and how it pairs with vector search.
Step-by-step: build a FastMCP server, deploy it to Cloud Run with IAM authentication, and connect an MCP client through the Cloud Run proxy.
Fourteen genuinely free AI courses, rechecked in November 2026 and grouped by level, from AI literacy for non-coders to building agents and MCP servers.
My short list of MCP servers worth installing for everyday work: Filesystem, Notion, Slack, Fetch and GitHub, with setup notes and the risks to watch.
Two ways to connect Claude Desktop to a remote MCP server, custom connectors or the mcp-remote bridge, with working config and troubleshooting.
No articles on this page match that.
The filter only looks at what is listed here. Search the whole site instead:
Search everything