Per-tool retry budget resets when the model interleaves another tool call, so an always-failing tool is retried past max_retries

The tool manager rebuilt each tool's retry counter every run step from only the tools that failed in that step, so a tool that was simply not called in an intervening step lost its accumulated count. A failing tool interleaved with a different tool call never reached its retries limit and could loop until the unrelated request limit.

Framework / agent

Pydantic AI · tool-calling agent

Remediation attempts

TESTED

Recurrence

not documented

Source languages

en

Updated

2026-09-30

Sources

• GITHUB ISSUE Tool `max_retries` budget silently resets if the failing tool isn't called in an intervening run step — github.com/pydantic/pydantic-ai, retrieved 2026-09-29
• GITHUB PULL REQUEST Fix tool retry counts across skipped steps — github.com/pydantic/pydantic-ai, retrieved 2026-09-29

Symptoms

• With retries=2, the failing tool is called and fails 6 times while ctx.retry stays at 0, and the run finishes without the max-retries error
• Against a model that keeps interleaving calls, the loop continues past the retry budget until request_limit (default 50)

Similarity to your system is not implied. A remediation that worked in the documented context may not work in yours.