ai exports

Tap. Paste.
Fixed.

Most debugging time isn't finding the bug — it's transcribing it. Sharingan's exports turn "can you describe what happened?" into a single structured paste that an AI agent (or a teammate) can act on immediately.

The workflow

Debugging with a user, a QA engineer, or your own device in hand:

That's the whole loop. The paste contains everything the agent needs to reason about the failure — no screenshots of JSON, no "what was the status code?" round trips.

What the agent receives

Structured Markdown with stable field labels, the body fenced and typed:

clipboard after "Copy for AI agent"
## HTTP 500 — GET /v1/streams/cam-12
host: api.iot-fleet.dev
duration: 10 214 ms (TTFB 10 180 · download 34)
request:
  Authorization: ••••
  Accept: application/json
response body:
```json
{ "error": "upstream_timeout",
  "stream": "cam-12", "retry_after": 30 }
```

Why this beats a HAR file or a screenshot:

Export API — every formatter is public

The share sheet has no private magic. Scripts, tests and agent harnesses can produce byte-identical output:

FunctionProduces
SharinganExport.agentMarkdown(event)One event as structured Markdown (the "Copy for AI agent" payload)
SharinganExport.agentMarkdown(events)Whole session with a counts header
SharinganExport.curl(httpEvent)Reproducible curl command — redactions stay masked
SharinganExport.json(event)Machine-readable single event
SharinganExport.sessionJson(events)Full session with tool metadata
SharinganExport.summary(events)Human digest, one line per event
e.g. attach a session to a bug report automatically
val report = SharinganExport.agentMarkdown(Sharingan.events.value)
bugTracker.attach("network-session.md", report)

AGENTS.md & llms.txt — for the agent's side

The repository ships a terse, complete API reference designed for machine consumption: AGENTS.md, mirrored at llms.txt. Point your coding agent at it and it knows the full capture, control and export surface without reading source.

Team setup that works well: add a line to your project's agent instructions — "Network/MQTT/BLE traffic is captured by Sharingan; ask the user to share 'Copy for AI agent' output, and consult its llms.txt for the API." Your agent then knows to request exactly the artifact Sharingan produces.

Privacy guarantees