Agent Run Inspector — Individual License

Turn an OpenCode session export into a readable incident report

Find repeated operations, recovered and unresolved failures, and unfinished calls in one local HTML report. Inspect one exported run at a time without uploading it to a service.

See the output before buying: Open the retry-loop sample report: https://kaboumou.github.io/agent-run-inspector-docs/samples/example-retry-loop-report.html — synthetic data, no signup or download required. Compare all four samples: https://kaboumou.github.io/agent-run-inspector-docs/.

Compatibility: OpenCode 1.18.x exports; Python 3.9+. Tested on Windows 11 / Python 3.11. macOS and Linux are not seller-tested. Other OpenCode versions run best-effort with a warning. Claude Code JSONL is not supported.

What the report answers

• Which identical operations repeated? Distinguishes failing retries, recovered retries and repeats after success that may be intentional.

• Which tool failures recovered, and which remain unresolved?

• Did the run finish? Missing evidence is shown as unknown.

• Which calls were slow or produced large outputs? Usage and cost are provider-reported, not invoice figures.

• What changes were recorded? Attached-file metadata, patch paths and context-compaction markers appear in the timeline.

Local workflow

Export a session, then generate its report:

opencode export ses_abc123 > session.json

runinspector session.json

The program uses the Python standard library, makes no network requests, and does not modify the input. Reports are deterministic and imported text is HTML-escaped. Reports can contain your session content: review them before sharing.

Included in version 0.1.3

• Python program source and installable wheel.

• README with installation, quickstart and troubleshooting.

• Four synthetic exports with HTML/JSON reports.

• Licence file. Optional JSON output for your own analysis.

Individual licence — USD 49 once

One person may use the purchased build internally. No redistribution or resale.

One-time purchase, no subscription. No installation service, customization, hosting, ongoing monitoring, guaranteed response time or guaranteed future updates. Future versions are optional separate purchases.

14-day refund on request. Purchase and delivery problems are handled under the published refund policy.

Scope

Inspector diagnoses recorded activity. It does not prevent loops, fix your agent, monitor live sessions or verify provider invoices. The free Agent Run Guard is a separate runtime tool; it is not required to use Inspector.

Download the free synthetic sample pack: https://payhip.com/b/I6OnS · Team licence: up to five named users — USD 149: https://payhip.com/b/w10Iu