An Agent Can't Offer an MCP Tool It Can't See
MCP discovery tells an agent how to authenticate, not what a tool does. When a tool sits behind per-user login, static metadata is the fix, not a smarter model.
~ % grep -r "agents" ./blog
3 posts found
MCP discovery tells an agent how to authenticate, not what a tool does. When a tool sits behind per-user login, static metadata is the fix, not a smarter model.
Hard limits for AI agents belong in code, not the prompt: the model reads prompt rules and breaks them. Plus the deny-by-default trap that does nothing.
AI made finding code issues free, but owning the fix still needs a human. Why finding throughput is a vanity metric and the hand-back is the real work.