HaiAI123

Curated Global AI Tools Directory

Industry update
Flat

A convention for returning tool-call failures to AI agents

11/15/2026 — 11/15, 05:15·1 sources (all)·1 reports (all)

Key takeaways

  • A developer published an article on DEV Community about how tool calls should report failures back to AI agents. The article separates failu…

Background & analysis

On November 15, 2026, a developer published an article on DEV Community examining how an Agent should receive errors when a tool call fails. The piece argues that a tool returning only the word "failed" is the worst possible outcome, because an Agent reading that single label cannot tell three entirely different situations apart. In the first, the request never reached the service at all, nothing happened, and retrying carries no cost. In the second, the request reached the service, the service refused the request, and the underlying result did not change, so retrying will fail again. In the third, the request reached the service but the final outcome is unknown. The author maintains that only the first situation is actually worth retrying.

To address the problem, the developer proposes a small set of conventions. Tools should return errors categorized by type, callers should attach an idempotency key, and a timeout budget should be defined. The author also recommends one unified rule for handling cases where the result is unknown, so that a retried call does not lead to a double charge or a duplicate message. Taken together, the piece presents a design-level proposal rather than a deployed standard. According to the available reporting, the discussion stops at the point where these conventions are put forward, and no follow-up on implementation, adoption, or agreement among maintainers is recorded yet.

AI-generated from 1 reports · updated 2 days ago

updream
24-hour heatHeat index 25 · peak 40 · 23h ago
24 hours agonow

Reports on this story headlines open the original

Nov 15
  1. A developer argues that returning a bare "failed" from a tool call leaves an agent unable to tell three very different situations apart: the request never reached the service, the service rejected it and nothing changed, or the outcome is unknown. Only the first is worth retrying. The proposed fix is a small contract: classified error returns, an idempotency key, a deadline budget, and one rule for what to do when the result is unknown.

    DEV Community · AIAI score 56

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to AI industry updates →