Urgent.News

What's breaking now, across thousands of outlets.

AI

Give your AI agent a way to say "I don't know": evidence envelopes over MCP

An AI agent is asked, "Is this French supplier still active, and what is its legal name?" It calls a tool, gets some text back, and replies confidently. What the user doesn't get is where the answer came from, when it was checked, or which parts the model filled in by itself. This tutorial is about a small pattern that closes that gap: instead of an answer, the tool returns an evidence envelope ,…

When an AI agent is asked a question such as "Is this French supplier still active, and what is its legal name?", it can generate a plausible answer even if it's incorrect. This is because language models predict plausible text based on patterns they've learned from training data. In the case of FACTRAIL MCP, a tool called when the agent is unsure, it returns an evidence envelope instead of a simple answer.

This envelope contains structured information about the facts the agent could establish and which ones remain unknown. The evidence envelope includes details like the overall outcome (supported, contradicted, insufficient evidence, stale, or conflicting sources), the specific facts, their support levels, the sources backing each fact, which requested fields were resolved and why, any competing claims, the freshness of the evidence, and a unique receipt ID for auditing purposes.

By providing this transparent breakdown of what is known and what remains unclear, the evidence envelope helps users understand the limitations of the AI's knowledge and encourages the model to acknowledge gaps in information rather than fabricating facts.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Building an AI Customer Support Agent with Memory

Customer-support systems can answer questions quickly, but a useful support agent needs more than a good-looking response. It needs to understand what happened before.

  • AI customer support agent developed with memory/retrieval system
  • Agent.py component incorporates retrieved memories into LLM prompts
  • Implemented guidelines to prevent AI from fabricating details

More from Tuesday 29 September →