Skip to content

fix(webui): honest server, connection and connect-modal states - #1053

Merged
github-actions[bot] merged 8 commits into
mainfrom
fix/ux-servers-connect
Aug 25, 2026
Merged

github-actions[bot] merged 8 commits into
mainfrom
fix/ux-servers-connect

Conversation

@Dumbris

@Dumbris Dumbris commented Aug 25, 2026 •

Copy link
Copy Markdown
Member

Fixes all 16 findings of the macOS tray UX + settings-parity audit (docs/qa/ux-audit-macos-tray-2026-08.md on audit/ux-macos-tray-sweep), in the sequence that audit suggests.

P1

  • F1 — the menu-bar icon now badges server health. Menu/TrayIcon.swift (a SwiftUI label nothing ever instantiated) is deleted along with AppState.healthLevel/HealthIndicator, and updateStatusIcon() draws the red/orange severity dot itself. It is an attributed-title glyph, not a composite image: the status image must stay isTemplate to follow the menu bar, and a template image is re-rendered monochrome.
  • F2 — every updateStatusIcon() pass sets the image's accessibilityDescription, the button's toolTip and its accessibility label ("MCPProxy — 13/29 servers, 942 tools — 5 server errors").
  • F3 — enable/disable/restart/login no longer swallow errors. One perform(_:on:id:) helper dispatches all of them and raises an NSAlert on failure, with Open Server Details / Report an Issue / Dismiss.
  • F4 — a "Needs Attention" row no longer runs an API action on a click that reads as navigation. Actioned rows became submenus carrying the verb (Sign in, Restart) plus Open Server Details; rows with no action stay plain navigate-only rows.
  • F5 — SidebarItem.tokens wires the finished TokensView into the main window.

P2

  • F6 — security.deep_scan.enabled, instructions (new .textarea control, placeholder fetched live from /api/v1/status.default_instructions) and code_execution_max_parallel added. scripts/check-settings-parity.py — already a required CI gate — was widened from duration fields to full key coverage plus placeholder/optional/step/min/max, excluding server_edition.*. It now pins 58 settings.
  • F7 — Enable/Disable unconditionally; stdio no longer gets Start/Stop (which contradicted its own "Disabled" status line).
  • F8 — (a) Review quarantine… in the per-server submenu, deep-linking to Server Detail → Tools. (b) is not in this PR: a native Security section wrapping deep scans, scan reports and TPA findings is the audit's own L-sized item and needs its own spec.
  • F9 — "Run at Startup" → Launch at Login (the name Settings already used), "Open MCPProxy..." → Open MCPProxy Window, "Open Web UI" → Open Web UI in Browser, plus a Settings footnote separating the two core-start controls from the menu's "Start/Stop MCPProxy Core".
  • F10 — a glance row's session id now reaches ActivityView over a new .activityFilter notification and seeds a session_id filter with its own dismissable chip.

P3

  • F11 — profiles read "deploy (2 servers · 27 tools)", and one whose servers are not in the config reads "research — no servers" with a tooltip naming them.
  • F12 — Protocol: is display-normalised (HTTP (streamable), HTTP (SSE), HTTP, stdio).
  • F13 — 512m / 1.0 / docker.io placeholders, step on entropy_threshold and oauth_expiry_warning_hours (with a stepper control), telemetry copy caught up to the Web UI. All now held by the parity gate.
  • F14 — ProjectLinks.issues/.discussions are used: Report an Issue and Discussions in the About panel, in Settings → App, and as a button on the F3 failure alert. The unreferenced openConfigFile() selector is deleted. Deviation from the audit's literal fix: not a new top-level menu row — fix(tray): slim the menu + represent code_execution by its sub-calls #1023 deliberately cut the menu to stop it scrolling and MenuStructureTests pins that; these homes cost no row.
  • F15 — the Servers submenu groups by state (needs attention → active → disabled) and folds 3+ disabled servers behind Disabled (N) ▸.
  • F16 — a native Tools sidebar item: the whole catalogue when the box is empty, BM25 ranking over /api/v1/index/search when it is not. (APIClient.searchTools was calling /api/v1/tools?q= — an endpoint with no q — and reading a field that endpoint never returns, so it had always returned []; nothing consumed it.)

Two new Swift suites (TrayPresentationTests, TrayAuditMenuTests, TrayAuditSurfacesTests) drive the real rebuildMenu() and the presentation rules; the suite is 1107 tests. No Go changes.

Cross-model review

Reviewed with opencode (gpt-5.6-sol), scoped to the Go seams. Every finding was verified against the source before being acted on.

Round 1 — three findings, all genuine, all fixed:

  1. edit_url could be offered to a server with no URL. The action was inferred from free-text last_error alone, and a stdio server emits those phrases from its own failed network calls — an npx/uvx install that cannot reach its package registry says no such host. HealthCalculatorInput gained HasEndpointURL; ActionEditURL now requires it. Verified live: a stdio server whose last_error contains no such host reads action = restart, summary still Host not found.
  2. The wrapper collapse treated every ": " as a wrapping boundary, so an adjacent repeat inside a quoted URL, JSON, captured stderr or validation text would have been deleted silently. A segment is now dropped only if it also looks like wrapper prose (short; letters, digits, spaces and word-joining marks only). Tests cover each payload shape the reviewer named.
  3. registered_url was documented as "never a credential", stronger than the shared sanitizer guarantees — Spec 091's sanitizeEndpoint keeps scheme, host and path. Reusing it is deliberate (the status row and the connect preview must not disagree about what an entry says), so the over-claim was the defect. Corrected in the Go comment, the TS type and the REST docs.

Reviewer also confirmed no content-read regression in GetAllStatus (Spec 075 FR-001 intact), that the new real-endpoint-over-name-only matcher precedence removes a previously randomized choice rather than introducing one, that oauth.IsOAuthError still sees the pre-collapse error, and no nil-deref/unhandled-error/concurrency defect.

Round 2 — one finding, genuine, fixed: the HasEndpointURL gate was correct in runtime.GetAllServers but missing from the other two production projections (internal/server/server.go, internal/server/mcp.go). The field defaults to false and false is fail-safe, so nothing broke loudly — the MCP upstream_servers list and the tray payload simply offered Restart where the Web UI offered Edit URL. Surfaces disagreeing about how to fix the same server is the exact defect class this PR exists to remove, and it is invisible to any single-surface test. Both call sites now set it, and a drift guard in the package that owns the contract fails if a file calling CalculateHealth does not assign the field (verified to fail both when the assignment is deleted and when it is replaced by a comment naming it).

Web-UI UX audit (docs/qa/ux-audit-webui-2026-08.md) findings F7, F10, F11,
F12, F17, F18, F28, F34 plus the F36 items in the same area.

F7 — a quarantined server card stated "needs security review" and afforded
no way to review. The card now carries a Review action deep-linked to the
server's Security tab (where the spec-088 banner holds the verdict and the
approve/scan actions), demotes Enable while the server is still held back,
and the greyed Scan button says why it is disabled.

F10 — the Dashboard "AI Agents" box read ClientStatus.connected from
GET /api/v1/connect, which is content-read-free by design (Spec 075) and so
reports connected=false for every client, always. The box could never say
"Connected" while /sessions listed a live session on the same data. It now
derives live clients from active MCP sessions — the same source /sessions
and the macOS tray use — with a last-seen time and a link to the session.
api.getSessions gains the backend's existing status filter so a small limit
cannot be filled by closed sessions and hide a live client.

F11 — a DNS failure offered "Restart", which redials the same broken
address forever. health.Action now resolves to a new "edit_url" for address
errors (no such host, unsupported/missing scheme, malformed URL);
"connection refused" deliberately keeps Restart. The action deep-links to
the Configuration tab with ?focus=endpoint, and the endpoint is editable
there. The server detail page also spoke three status vocabularies at once
(Enabled/Active, Online/Offline, Disconnected) with a green check on a
server that was down; the tiles now carry exactly two axes — the admin
state you set and the health level we observe — and the header badge uses
the same unified reading as the Servers-page card.

F12 — the Servers card printed the full wrapped Go error chain beside a
perfectly good badge. The card shows the plain-language summary with the
raw chain behind a details disclosure. The duplicated wrapper text is fixed
at the source: mcp-go's streamable-HTTP transport wraps a send failure with
the same message at two nesting levels, so adjacent duplicate segments are
collapsed once, centrally, on last_error and on the diagnostic cause.

F17 — the three registry cards sat above the selector looking exactly like
one and were not clickable. They are the selector now (click or keyboard to
toggle, visible selected state), the kebab menu no longer doubles as a
selection, and the disabled search box says what would enable it.

F18 — "connected" meant "the config holds an entry named mcpproxy", so a
client pointing at a different instance on a different port showed as
connected here. ClientStatus gains proxy_url, registered_url (sanitized the
same way a preview's entry summary is) and endpoint_match (this/other/
unknown); a row addressing another instance says so and offers "Point here"
instead of a false Disconnect. All rows use one action pattern, Disconnect
confirms first and states the backup up front, Connect All names how many
clients it would change and explains itself when it is off, and a client
with no config names the paths that were searched.

F28 — three surfaces for one cause: the auth modal, an inline load error
behind it, and a "Connection Lost" toast. The modal now owns the message;
the key field is labelled required rather than optional; "Continue Without
Auth" is demoted and honest about leaving pages empty; and the key-recovery
instructions no longer assume a tray, which headless and server installs do
not have.

F34 — the security tab printed the Risk Score twice and a bare coloured
gauge with an unexplained number. One score, a labelled gauge, and a
copyable scan id.

F36 — the hub read "just started" on every page load because uptime was
measured from when the page first saw the core running; /status and the SSE
status frames now report the core's own started_at. The Sessions table's
"View Activity" label no longer wraps onto two clipped lines.
Live verification of F12 turned up a defect in the fix itself: health.summary
is only the mapped error phrase while a server is administratively enabled.
For a disabled or quarantined server the calculator short-circuits and the
summary describes the admin state, so the card's red alert read "Quarantined
for review" — restating the quarantine banner directly below it instead of
saying why the connection failed. Fall through to the structured diagnostic's
user message, then to the raw chain's root cause.

Also gives the server-detail Admin state tile a power symbol rather than the
same gear the Tools tile uses: it is the on/off decision you made, not a
configuration detail.
@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Aug 25, 2026 •

Copy link
Copy Markdown

Deploying mcpproxy-docs with  Cloudflare Pages  Cloudflare Pages

Latest commit: 428b435
Status: ✅  Deploy successful!
Preview URL: https://5f61f17f.mcpproxy-docs.pages.dev
Branch Preview URL: https://fix-ux-servers-connect.mcpproxy-docs.pages.dev

View logs

… inputs

Cross-model review (opencode / gpt-5.6-sol) raised three findings on the
backend seams. All three verified; all three addressed.

1. edit_url could be offered to a server with no URL to edit. The action was
   inferred from free-text last_error alone, and a stdio server emits exactly
   those phrases when its own network calls fail — an npx/uvx install that
   cannot reach its package registry says "no such host". HealthCalculatorInput
   gains HasEndpointURL, set from the configured URL, and ActionEditURL now
   requires it. Restart remains the remedy for stdio.

2. The wrapper collapse treated every ": " as a Go wrapping boundary, so an
   adjacent repeat inside a quoted URL, a JSON fragment, captured stderr or
   validation text would have been deleted silently. A segment is now dropped
   only when it also looks like wrapper prose: short, and made solely of
   letters, digits, spaces and word-joining marks. Anything carrying a quote,
   brace, bracket, slash or equals sign is payload and survives verbatim, even
   when it repeats. Tests cover each of those shapes.

3. registered_url was documented as "never a credential", which is stronger
   than the shared sanitizer guarantees: Spec 091's sanitizeEndpoint keeps
   scheme, host AND path, so a credential embedded in a path segment by a
   third-party config would survive — exactly as it already does on the connect
   preview surface. Reusing that sanitizer is deliberate, so the status row and
   the preview cannot disagree about what an entry says; the over-claim, not
   the behaviour, was the defect. Corrected in the Go comment, the TS type and
   the REST docs.

Reviewer also confirmed no content-read regression in GetAllStatus (Spec 075
FR-001 intact), that the new real-endpoint-over-name-only matcher precedence
removes a previously randomized choice rather than introducing one, that
oauth.IsOAuthError still sees the pre-collapse error, and no nil-deref,
unhandled error or concurrency defect.
…site

Round-2 cross-model review found the HasEndpointURL gate was correct in
runtime.GetAllServers but missing from the other two production projections
(internal/server/server.go and internal/server/mcp.go). The field defaults to
false and false is fail-safe, so nothing broke loudly — a URL-addressed server
was simply offered "Restart" on the MCP `upstream_servers` list and the tray
payload while the Web UI offered "Edit URL". Surfaces disagreeing about how to
fix the same server is precisely the class of defect this change set exists to
remove, and it is invisible to any single-surface test.

Both call sites now set it from the configured URL, and a drift guard in the
package that owns the contract asserts that any file calling CalculateHealth
also mentions HasEndpointURL — verified to fail when the field is removed.
The drift guard matched the bare identifier, so a passing comment naming
HasEndpointURL would have satisfied it — exactly the vacuous green the test
exists to prevent. It now requires an actual assignment (a struct-literal
field or a field write). Verified to fail when the assignment is replaced by
a comment mentioning the field.
…view config

internal/server/server.go already resolves the server's config fields through
a storage fallback: when the stateview carries no Config it re-reads the stored
upstream and fills `url` from that. My previous commit read
serverStatus.Config.URL directly, so on exactly that fallback path an HTTP
server would report HasEndpointURL=false and be offered Restart instead of
Edit URL — reintroducing the surface disagreement one branch lower down.

Use the resolved `url`. OAuthRequired is left reading serverStatus.Config
exactly as before; changing its source is a separate question.

internal/server/mcp.go needs no change: its `server` IS the config struct, with
no fallback path.
@github-actions

github-actions Bot commented Aug 25, 2026 •

Copy link
Copy Markdown
Contributor

📦 Build Artifacts

Workflow Run: View Run
Branch: fix/ux-servers-connect

Available Artifacts

  • archive-darwin-amd64 (29 MB)
  • archive-darwin-arm64 (26 MB)
  • archive-linux-amd64 (17 MB)
  • archive-linux-arm64 (15 MB)
  • archive-windows-amd64 (29 MB)
  • archive-windows-arm64 (25 MB)
  • frontend-dist-pr (0 MB)
  • installer-dmg-darwin-amd64 (23 MB)
  • installer-dmg-darwin-arm64 (21 MB)

How to Download

Option 1: GitHub Web UI (easiest)

  1. Go to the workflow run page linked above
  2. Scroll to the bottom "Artifacts" section
  3. Click on the artifact you want to download

Option 2: GitHub CLI

gh run download 32877931041 --repo smart-mcp-proxy/mcpproxy-go

Note: Artifacts expire in 14 days.

@codecov-commenter

codecov-commenter commented Aug 25, 2026 •

Copy link
Copy Markdown

⚠️ Please install the 'codecov app svg image' to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 70.75472% with 31 lines in your changes missing coverage. Please review.

Files with missing lines Patch % Lines
internal/connect/connect.go 70.83% 13 Missing and 1 partial ⚠️
internal/runtime/runtime.go 0.00% 11 Missing ⚠️
internal/health/calculator.go 88.88% 1 Missing and 1 partial ⚠️
internal/httpapi/server.go 33.33% 2 Missing ⚠️
internal/management/service.go 0.00% 1 Missing ⚠️
internal/server/server.go 0.00% 1 Missing ⚠️

📢 Thoughts on this report? Let us know!

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved (Model B): Paperclip review verdicts = ACCEPT and qa-gate green at this head SHA. Arming auto-merge; GitHub merges when all required checks pass.

@github-actions
github-actions Bot merged commit 4974fb6 into main Aug 25, 2026
41 checks passed
Dumbris added a commit that referenced this pull request Aug 26, 2026
The document read as an open P1–P3 backlog, but #1053 (merged 2026-08-25) fixed
all 16 findings. Two readers in review took the header's "no version skew, so
every finding below is reproducible from source" at face value against current
main, where it is no longer true.

- Header now scopes that claim to the time of the walk and carries a status
  banner pointing at #1053.
- New §6 Resolution records the two fixes whose obvious implementation is wrong
  (F1's severity dot must be an attributed-title glyph, not a composite image,
  because the status image stays isTemplate; F6 added a parity gate, not just
  the three missing keys).
- The caveat is stated rather than hidden: this pass could not re-walk the live
  tray, because macOS Accessibility was revoked on the reviewing machine
  (check_accessibility -> trusted:false), so the resolution is read from #1053's
  diff, not from a second walk.
Dumbris added a commit that referenced this pull request Aug 26, 2026
The document read as an open backlog, but #1055 ("fix(tray): close the 16
findings of the macOS tray UX + settings-parity audit", merged 2026-08-25) fixed
all 16, with #1056 as a catalogue follow-up. The header's "no version skew, so
every finding below is reproducible from source" is no longer true against main.

- Header scopes that claim to the time of the walk and carries a status banner.
- New §6 Resolution records the two fixes whose obvious implementation is wrong:
  F1's severity dot is an attributed-title glyph, not a composite image (the
  status image stays isTemplate and a template image renders monochrome), and F6
  added a parity gate rather than just the three missing keys.
- Verified on main rather than taken on trust: Menu/TrayIcon.swift is deleted,
  TokensView is reachable from Views/MainWindow.swift, MCPProxyApp.swift sets
  isTemplate + accessibilityDescription + toolTip + attributedTitle glyph, and
  SettingsCatalog.swift carries deep_scan / instructions /
  code_execution_max_parallel.

Note for anyone tracing this later: #1053's body claims to fix these 16 tray
findings, but #1053 touches no Swift file — it is the Web UI cluster its own
title names. #1055 is the tray PR.

The caveat is stated rather than hidden: this pass could not re-walk the live
tray, because macOS Accessibility was revoked on the reviewing machine
(check_accessibility -> trusted:false), so the resolution is read from the #1055
diff, not from a second walk.
Dumbris added a commit that referenced this pull request Aug 26, 2026
The document read as an open P0-P3 backlog, but seven PRs merged on 2026-08-25
(#1044, #1048, #1049, #1051, #1052, #1053, #1054) closed it. Anyone opening this
file against current main was being handed 36 stale defects.

Added a status banner and a Resolution section that separates three things
honestly:

- Re-checked live on v0.61.0 against a running isolated core with real /mcp
  traffic: F1, F4, F5, F7, F8, F12, F13, F20, F24, F26, F33 confirmed fixed,
  each with the observed evidence rather than a PR reference.
- Still open: F6 is half fixed. Focus enters the Add Server dialog and submit is
  in a sticky footer, but Escape still does not close it - AddServerModal.vue has
  no key handler and JsonViewer.vue is the only component under components/ that
  handles Escape at all. F28's "key called optional" half is fixed; its
  three-surfaces state is not.
- Not re-checked: named explicitly rather than implied, so the next reader knows
  which closures rest on a PR body and which on an observation.

Five defects found on main while re-checking are recorded separately, since they
are not regressions of anything in the audit. The first is security-relevant:
quarantining a server through the config file leaves its tools in the search
index, so retrieve_tools still hands the agent their names and descriptions -
which is exactly what quarantine exists to prevent - before the call is refused.
Runtime.QuarantineServer purges the index; the reload path detects the same
transition (quarantined_changed: true) and does not.

Also corrects the record on #1053: its body names F1-F16, which is the tray
audit's numbering mis-pasted. #1055 is the tray PR; #1053 touches no Swift file.
Dumbris added a commit that referenced this pull request Aug 26, 2026
The document read as an open P0-P3 backlog, but eight PRs merged on 2026-08-25
(#1044, #1048, #1049, #1050, #1051, #1052, #1053, #1054) closed it. Anyone
opening this file against current main was being handed 36 stale defects.

Added a status banner and a Resolution section that separates four things:

- Re-checked live on v0.61.0 against a running isolated core with real /mcp
  traffic: F1, F4, F5, F6, F7, F8, F12, F13, F20, F24, F26, F33 confirmed fixed,
  each with the observed evidence rather than a PR reference.
- Not re-checked: named explicitly rather than implied, so the next reader knows
  which closures rest on a PR body and which on an observation. F28 is called
  out as partial - its "key called optional" half is fixed, its
  three-surfaces-at-once state is not.
- Corrections to the audit itself: F23 mis-diagnosed a per-request structural
  estimate as a stale aggregation, and F31's "Manual badge has no tooltip" half
  was wrong even at audit time (spec-088 trust-mode badge, already titled).
- Five defects found on main while re-checking, recorded separately since they
  are not regressions of anything in the audit. The first is security-relevant:
  quarantining a server through the config file leaves its tools in the search
  index, so retrieve_tools still hands the agent their names and descriptions -
  which is exactly what quarantine exists to prevent - before the call is
  refused. Runtime.QuarantineServer purges the index; the reload path detects
  the same transition (quarantined_changed: true) and does not.

F6 carries a methodology note. Re-checking it produced a false negative: Escape
driven through the browser-automation layer never reaches the page's document
keydown listener, so the dialog looked stuck open. Dispatched in-page it closes
at once, and #1050's useModalA11y is wired into all five <dialog :open> modals.
The sweep assertion this document proposes must dispatch in-page, or it tests
the automation layer instead of the app.

Also corrects the record on #1053: its body names F1-F16, which is the tray
audit's numbering mis-pasted. #1055 is the tray PR; #1053 touches no Swift file.
Dumbris added a commit that referenced this pull request Aug 26, 2026
* docs(qa): Web UI UX audit — ranked findings (ux-audit epic)

Heuristic + live-instance audit of every routed Web UI view in light/dark,
first-run and populated states, at 1440/820/390px. 36 findings ranked P0-P3
with view, evidence, heuristic violated, suggested fix and effort.

Audited against standard heuristics plus the house activity-UI rules
(grouping, failure-only status, intent reason, histogram/log parity).

* docs(qa): record what the Web UI audit's backlog actually resolved to

The document read as an open P0-P3 backlog, but eight PRs merged on 2026-08-25
(#1044, #1048, #1049, #1050, #1051, #1052, #1053, #1054) closed it. Anyone
opening this file against current main was being handed 36 stale defects.

Added a status banner and a Resolution section that separates four things:

- Re-checked live on v0.61.0 against a running isolated core with real /mcp
  traffic: F1, F4, F5, F6, F7, F8, F12, F13, F20, F24, F26, F33 confirmed fixed,
  each with the observed evidence rather than a PR reference.
- Not re-checked: named explicitly rather than implied, so the next reader knows
  which closures rest on a PR body and which on an observation. F28 is called
  out as partial - its "key called optional" half is fixed, its
  three-surfaces-at-once state is not.
- Corrections to the audit itself: F23 mis-diagnosed a per-request structural
  estimate as a stale aggregation, and F31's "Manual badge has no tooltip" half
  was wrong even at audit time (spec-088 trust-mode badge, already titled).
- Five defects found on main while re-checking, recorded separately since they
  are not regressions of anything in the audit. The first is security-relevant:
  quarantining a server through the config file leaves its tools in the search
  index, so retrieve_tools still hands the agent their names and descriptions -
  which is exactly what quarantine exists to prevent - before the call is
  refused. Runtime.QuarantineServer purges the index; the reload path detects
  the same transition (quarantined_changed: true) and does not.

F6 carries a methodology note. Re-checking it produced a false negative: Escape
driven through the browser-automation layer never reaches the page's document
keydown listener, so the dialog looked stuck open. Dispatched in-page it closes
at once, and #1050's useModalA11y is wired into all five <dialog :open> modals.
The sweep assertion this document proposes must dispatch in-page, or it tests
the automation layer instead of the app.

Also corrects the record on #1053: its body names F1-F16, which is the tray
audit's numbering mis-pasted. #1055 is the tray PR; #1053 touches no Swift file.
Dumbris added a commit that referenced this pull request Aug 27, 2026
* docs(qa): macOS tray UX + settings-parity audit

Read-only sweep of the running v0.60.0 tray (accessibility-tree menu walk)
plus a static comparison of the tray's native surfaces and settings
catalogue against the Web UI.

Parity matrix (feature surfaces, settings fields, per-server actions) and
16 ranked findings: dead menu-bar health badge, unlabelled status item,
silent action failures, unreachable Agent Tokens view, and three settings
fields that drifted out of the tray catalogue.

* docs(qa): mark the tray audit resolved, and say what was not re-verified

The document read as an open backlog, but #1055 ("fix(tray): close the 16
findings of the macOS tray UX + settings-parity audit", merged 2026-08-25) fixed
all 16, with #1056 as a catalogue follow-up. The header's "no version skew, so
every finding below is reproducible from source" is no longer true against main.

- Header scopes that claim to the time of the walk and carries a status banner.
- New §6 Resolution records the two fixes whose obvious implementation is wrong:
  F1's severity dot is an attributed-title glyph, not a composite image (the
  status image stays isTemplate and a template image renders monochrome), and F6
  added a parity gate rather than just the three missing keys.
- Verified on main rather than taken on trust: Menu/TrayIcon.swift is deleted,
  TokensView is reachable from Views/MainWindow.swift, MCPProxyApp.swift sets
  isTemplate + accessibilityDescription + toolTip + attributedTitle glyph, and
  SettingsCatalog.swift carries deep_scan / instructions /
  code_execution_max_parallel.

Note for anyone tracing this later: #1053's body claims to fix these 16 tray
findings, but #1053 touches no Swift file — it is the Web UI cluster its own
title names. #1055 is the tray PR.

The caveat is stated rather than hidden: this pass could not re-walk the live
tray, because macOS Accessibility was revoked on the reviewing machine
(check_accessibility -> trusted:false), so the resolution is read from the #1055
diff, not from a second walk.

* docs(qa): correct two §1 claims that were wrong at the time of the walk

Verified against v0.60.0 rather than taken on trust:

- The tray Settings tab is named "Raw", not "Raw JSON"
  (native/macos/MCPProxy/MCPProxy/Views/SettingsView.swift:38 at v0.60.0). The
  §1.2 inventory, the §2 parity matrix and F6's body all said "Raw JSON".
- The Web UI had no /usage route at v0.60.0 - the router's path list runs
  /, /servers, /repositories, /search, /settings, /secrets, /sessions, /tools,
  /activity, /security, /tokens with no usage or overview entry. Usage was a
  Dashboard tab; /usage became a route later, in #1044. The parity matrix's
  "Usage report | ⚠ savings % on Dashboard | ✅ /usage | inconsistent" overstated
  the Web UI side of that comparison.

Also noted in §6 that three §1.1 menu labels have changed since the walk as part
of the F9 fix: Open MCPProxy Window, Open Web UI in Browser, Launch at Login.

* docs(qa): re-walk the live tray and record that the findings still hold

Replaces the "could not re-walk" caveat with the walk itself. macOS
Accessibility was granted, so mcpproxy-ui-test works now; the walk reproduces
this audit almost verbatim, because the installed tray is still v0.60.0 with
"Update 0.61.0 — ready to restart?" sitting in the menu.

Reproduced live: F2 (empty status-bar title, no description, no tooltip), F5 (no
Agent Tokens entry), F7 (ElevenLabs "Stop" vs cloudflare-graphql "Disable"; and
memory showing status Disabled above action Start), F8 (all three quarantined
servers offer only Disable/Stop, Restart, View Logs), F11 (research and deploy
still offered at 0 tools), F12 (streamable-http beside http and sse), F15 (29
flat entries), F16 (no tool search), and the pre-#1055 menu labels.

One drift worth noting: the Clients block now reads "No clients in the last 24h"
rather than listing an idle client - the empty state H5 exists to avoid.

The consequence for the resolution section: #1055's fixes are real in the diff
but remain unverified against a running tray. That needs the pending 0.61.0
restart or the dev bundle-swap, neither of which this pass did - restarting the
tray kills the core it manages.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants