runlens is a Go CLI for analyzing JSONL traces from agent runs, tool calls, and evaluation loops.
It helps you spot failures, flaky paths, latency outliers, and token-heavy runs before they turn into production issues.
brew install itamaker/tap/runlensYou can also download binaries from GitHub Releases.
Current release archives:
- macOS (Apple Silicon/arm64):
runlens_0.2.0_darwin_arm64.tar.gz - macOS (Intel/x86_64):
runlens_0.2.0_darwin_amd64.tar.gz - Linux (arm64):
runlens_0.2.0_linux_arm64.tar.gz - Linux (x86_64):
runlens_0.2.0_linux_amd64.tar.gz
Each archive contains a single executable: runlens.
Run:
runlensThis launches the interactive Bubble Tea terminal UI.
You can still use the direct command form:
runlens summary -input examples/run.jsonlFor a deeper diagnosis:
runlens diagnose -input examples/run.jsonl- Go
1.22+
go run .Or run a direct command:
go run . summary -input examples/run.jsonlMachine-readable output:
go run . summary -input examples/run.jsonl -jsonDiagnostic report:
go run . diagnose -input examples/run.jsonlmake buildgo build -o dist/runlens .- Parses JSONL event streams from agent or tool executions.
- Computes aggregate latency, success rate, and token totals.
- Produces per-tool summaries for failure analysis.
- Clusters recurring errors, surfaces flaky tools, and highlights latency outliers.
- Exports either human-readable output or JSON for automation.
examples/run.jsonlis a good shape reference for your own logs.- Maintainer release steps live in
PUBLISHING.md.
 |
|---|
| Zhaoyang Jia |
