Skip to content

Free trace audit

Upload one trace at https://slancha.ai/audit. See findings in your browser and get a report by email. No sign-in or paid plan.

What you get

  • Recorded error patterns, model workload and counts of likely secrets.
  • Today, the audit uses fixed rules; no language model reads the file.
  • Findings are observations. They do not show a root cause or savings.

Before you upload

  • Use a work email. Confirm you have the right to upload the data.
  • Accepted: an observability trace export (JSONL or JSON), a Claude Code transcript (.jsonl) or an OpenAI chat-completion log.
  • Size limit: 4,000,000 bytes after browser compression.
  • Limits: 3 audits a day per email and per source IP, and 10 a day per company. A rejected file still counts.
  • Errors are classified for a Claude Code transcript of one session only.
  • Do not upload secrets you cannot rotate. The audit counts likely secrets but does not remove them.

What happens to your file

  • The browser compresses your raw file and sends it to Slancha's audit service. Today, the page does not send the file name.
  • Today, the service writes the file to disk with owner-only permissions while it analyses it.
  • This is separate from optional workspace tracing. Tracing is a different flow that a workspace admin turns on for Claude Code sessions. The audit does not use it, and uploading to the audit does not turn it on.
  • No account is created.
  • Today, Slancha records email and source-IP digests, email domain, time, size, format and outcome for each audit. Retention periods are not set out in this guide.
  • Today, analytics record funnel steps under a random flow ID. Do Not Track turns off analytics and email engagement tracking.
  • Email open and click tracking applies. Privacy tools can trigger opens automatically.
  • Today, requesting an audit does not enroll you in a follow-up sequence; that needs a separate choice.
  • Privacy questions: privacy@slancha.ai.

The email

We email a summary, the full safe report as an HTML attachment and supplementary JSON. Today, the report email includes findings, aggregate resource metrics and coverage. It does not include trace text, identifiers, names, locations, timestamps or secret values. See the audit privacy notice.

If something fails

  • Audit failed or the service is busy ("Couldn't reach the audit service, or it's busy", "The audit service is busy"): the page suggests trying again in a minute. This is a new upload. It is not one of the two UI retries, and a rejected or repeated file can use another audit from your daily limit.
  • Connection ended before a result ("couldn't confirm the audit result"): check your inbox first. The server may still finish and email you. Another upload may use another audit.
  • The page waits up to 300 seconds. When the browser gives up after a transport failure, that does not show the backend stopped.
  • Email delivery failed: the page offers up to two UI retries of the email only. A retry appears only when the report says it supports retry. It resends the already-issued safe report. It does not upload your trace again or create another audit.
  • Email delivery unknown: check your inbox and spam folder before any manual retry, because the server may have sent it. You can recover a local safe copy from the page. There is no automatic replay and no need to upload again.
  • Safe report unavailable ("Safe report unavailable"): the audit completed but the safe report could not be prepared. The page offers no download and no resend. Keep your original file and contact Slancha with the audit ID. Do not upload it again just to retry delivery.

Developers

The upload is a public web form. There is no supported upload API.

PreviousDecisions pilot guideNextOptimize agents