JoesProxy feature catalog
Find the workflow for the question in front of you, from the first routed request to a repeatable test or a shareable result.
Connect a target
Choose the app making the request: Mac, Simulator, iPhone or iPad, or Apple TV. Follow the setup checkpoints to distinguish a reachable listener from a routed app. A CONNECT tunnel can answer “did it contact that host?” without decrypting the response. For authorized inspection, check trust and permissions; a browser passing verification does not prove another app accepts the certificate.
Read a capture
Start with a specific symptom: Why did a catalog refresh leave the list empty?
| What you want to know | Where to look |
|---|---|
| Which endpoint handled my action? | Use Table or Explorer to select a request and inspect its headers, body availability and response. |
| Which call came first? | Use Timeline and waterfall export to correlate lanes and gaps. Keep the chart’s qualifications when sharing it. |
| What survived after I closed the workspace? | Recordings, bookmarks, replay and HAR serve different purposes. Live retention limits what can be saved; reopening a file is not replaying the request. |
| Can I keep the evidence from an interrupted run? | Review the recovery offer before starting fresh. |
Change one condition and retry
Two worked examples: serve an empty catalog without changing the account, or verify a subtitle switch against a known HLS cue.
- Replace, redirect or block a response: Local overrides compares Map Local, Map Remote, blocking and live JSON rewrites. Start with the narrowest matching endpoint.
- Keep the input fixed: Bookmark response rewrites serve a patched saved snapshot. Use this when a changing upstream response would spoil the comparison.
- Coordinate catalog data and artwork: Trace an image, then package related rules and assets so another investigator can repeat the test.
- Change delivery conditions: Network conditioning adds bandwidth or latency constraints on proxied traffic. A long waterfall alone doesn’t prove a throttle was active.
- Investigate a stream: Media investigation covers HLS and DASH manifests, ad XML and controlled HLS fixtures. Check what the player requested separately from what appeared on screen.
Work with an agent or a script
Enroll an MCP client or set up the CLI. Use an agent investigation to find the relevant request, then analyze it on the Mac. For counts and totals, calculate over selected sessions rather than sending the whole payload. Compare workflow receipts to see whether this actually reduced calls, bytes or repeated work.
For exact names, examples, permissions and cleanup, browse the MCP tool, resource and prompt index or CLI command index. If an expected observation is missing, start with the symptom guide.