How-to

Reading audit results

Reading audit results ​

The card of a completed audit shows the outcome for the whole server, the outcome of each utility, and each check with its details. The first tab answers "what to fix", the utility tabs show why.

Role: a team member who can see the project.

Prerequisites ​

The audit has run and its status is Completed. MCP server audit run

Steps ​

  1. Open a project. In the project menu choose MCP Testing > Audits.
  2. Open the audit. The card title is ID and the audit number.
  3. On the Summary tab look at the tiles and the What to fix block.
  4. Click a check in What to fix. Its utility tab opens with the check card expanded.
  5. Read the description, the result details, and Advice.
  6. Fix the server and click Run again to get a new audit.

Audit card ​

Breadcrumbs: projects, project, MCP testing, server, Audits. Header buttons:

ButtonWhen shownWhat it does
RefreshalwaysReads the audit again. The card does not refresh by itself
CancelPending, In progressCancels the audit
Run againCompleted, Cancelled, ErrorCreates a new audit of the same server
ExportCompletedOpens the export dialog. Audit export
DeletealwaysDeletes the audit after confirmation

Tabs: Summary and one tab per utility in audit order. Utility tabs exist only on an audit with status Completed. Tabs that do not fit the width scroll with arrows.

Summary tab ​

Until the audit finishes, a message about its state stands in place of the outcome. On an Error audit the message quotes the cause. A Cancelled audit has no results.

A completed audit shows a message in two cases: some utilities ended with an error, or no check could be verified. In the second case make sure the server address is reachable from the internet and the access is valid, then run the audit again.

Tiles ​

TileWhat it counts
ViolationsChecks with status Failed at the Requirement level: what the MCP specification requires (MUST, MUST NOT)
RemarksChecks with status Failed at the Recommendation (SHOULD) and Consistency levels
AdviceChecks with status Failed at the Practice level: industry recommendations beyond the specification
Not verifiedChecks with status Not verified. This is not a violation: the network, a time limit, an unreadable answer, or a missing prerequisite stopped the audit
ChecksAll checks of the completed utilities in every status

The counters are the sum over the utilities with status Completed. The tile tooltip repeats the counting rule.

Results by utility ​

One row per utility: status, a bar of checks by status, violations, remarks, advice, and the number of checks. An Error utility shows "Checks did not run" in place of the bar, a Not applicable utility shows "Does not apply".

What to fix ​

All checks with status Failed from all utilities. Sorted by norm level: requirement, then recommendation and consistency, then practice. Inside a level, in audit order. Columns: norm level, check, utility, one-line result, status. A check opens its card in the utility tab.

Not verified ​

Checks with status Not verified, grouped by cause. One cause often covers many checks: for example, "Time limit exceeded" on one request. The causes are listed in the reference. Check result detail

Details ​

A side block: status, server, author, creation, start and finish time, duration, check catalog version, runner version. An Error audit also shows the failure reason.

Key measurements ​

A side block on a completed audit when the server answered: protocol epoch, transport, number of tools, whole tool definition in tokens, share of the context window, days until the certificate expires, and the certificate end date. The window share and the certificate term have a bar. The certificate scale is 90 days.

Utility tab ​

The top bar has a status filter with counters (All and each status), Expand all, and Collapse all. The filter only hides cards, the check order does not change. The check order is always the same, as in the catalog.

The side Details block of the utility shows its status. A Completed utility also shows violations, remarks, advice, the number of checks, a bar, and the number of checks per status. An Error utility shows the failure reason.

An Error or Not applicable utility shows a message in place of checks. An error in one utility does not stop the others.

Check card ​

The header holds the check name, the one-line result, the norm level, and the status. Cards with status Failed are expanded at once.

An expanded card shows:

  1. The description: what the check establishes.
  2. The result details: what was received, what was expected, what was measured. The layout depends on the data: one line, "name - value", chips, a table, or a table of objects with the field path. When there is an expected value, the Received and Expected blocks stand side by side. A table shows the first 8 rows, Show all opens the rest.
  3. Advice: how to fix it. Advice is shown only on a check with status Failed.

Numbers carry units: milliseconds, bytes, kilobytes and megabytes, percent. Codes and HTTP statuses have no digit grouping. A value the server did not report is labeled "not reported".

Card or Markdown ​

Each block (summary sections, side blocks, check cards) has a Card / Markdown switch and a Copy button in its header. Copy always puts the whole Markdown of the block on the clipboard, whatever view is open. The Markdown of a check holds the check key, norm level, status, result, description, details, and advice. Such a block can go to a language model or into a task. The chosen view is not stored.

Reading statuses ​

Check statusMeaning
PassedThe server meets the check
FailedThe server does not meet the check. Advice is available
Not verifiedThe audit did not get an answer it could judge. The cause is in the details
Not applicableThe check does not concern this server: for example, the server declares no prompts. The cause is in the details
MeasuredInformation without a verdict: a number, a list, a version

The full lists of statuses and norm levels are in Audit check.

Version 1.0.0 · Last updated 2026-11-13

Did this page help?