tool_error is one of Tessary's built-in classifiers. It watches the rate at which each tool's calls fail rather than judging individual failing calls. A call counts as failed on structural evidence only: an error status on the span, a recorded exception, or a result that itself declares an error. Output text that merely mentions the word "error" never counts.
It catches a tool going bad while the agent keeps running: an API that started refusing, a dependency that broke, a rate limit that began to bite. Each tool's alert threshold comes from its own baseline failure rate and an explicit false-alarm budget, and evidence accumulates across calls, so a tool failing at its normal rate stays quiet and only a sustained shift fires. A detection is one finding per tool: when the shift began, which failure patterns moved the rate, and which call sites the traffic came through. If triage confirms a real deviation, the finding opens a case.
5 questions
Answered, plainly.
Two ways to run Tessary.
Tessary is an open-source agent reliability platform. Cloud and self-hosted run the same workflow on the OpenTelemetry traces your agent already emits.
Tessary Cloud
We host it for you. Send your first trace with nothing to deploy and no model key.
- traces
- 10,000 per calendar month
- stored trace data
- 1 GB
- retention
- 30 days
- model credit
- $10, one-time, for triage and root-cause analysis
- credit card
- not required
Self-hosted Tessary
Run the open-source code on your own infrastructure with one command. Add your own model key for triage and root-cause analysis.
Self-host Tessary for me by following https://github.com/tessaryai/tessary/blob/main/setup.md
docker compose -f oci://docker.io/tessaryai/tessary:compose up -d -y