Reasoning shown, not a black-box score

Point it at your network.
Get told why it's slow.

A one-shot, no-UI CLI that captures live traffic and figures out whether your slowdown is your ISP, your own device, or one flaky destination — with the reasoning shown, every time.

Buy report credits — from $5 View source

$ npm install -g e3d-netdoctor

paid-report --interface en0 --card

$ e3d-netdoctor paid-report --interface en0 --card

Only analyze traffic on networks you are authorized to monitor.
Requesting e3d payment before capture/analysis...
✓ payment confirmed — 500 credits
{
  "findings": {
    "verdict": {
      "headline": "Likely upstream/ISP",
      "confidence": "High",
      "rationale": "7 of 8 independent
        destinations affected, spanning 3 providers"
    }
  },
  "payment": { "creditsRemaining": 1500 }
}

Why this exists

“Is it my wifi, or is it Comcast again?”
— every network engineer and Wireshark user, at some point

That question is answerable from a single packet capture — if you know to group retransmissions and RTT outliers by destination and by local device, and weigh provider diversity instead of raw destination count. Most people don't do that math by hand. netdoctor does, every time, and shows its work.

Likely Upstream/ISP

Signal spread across many unrelated destinations and providers — not one device.

Likely Local

Signal concentrated on one local device across many destinations — bad wifi, flaky NIC, bad cabling.

Likely Destination/Path-Specific

Signal confined to one or two destinations while everything else stays clean.

FOR AGENTS

Got an AI with shell access? Hand it the URL.

netdoctor doesn't need a human driving it. Every command is one call in, one structured JSON result out — no prose to re-parse, no dashboard to click through, no login. Give your agent netdoctor.e3d.ai and it can read llms.txt, install the CLI, capture, diagnose, and even pay for its own report — all without you touching a terminal.

→ Deterministic, rule-based verdicts — never another model call your agent has to trust blindly.
→ Machine-readable rationale, quotable straight back to whoever it's working for.
→ Non-interactive payment path (a pre-provisioned credit key) — safe to run on a schedule with zero human involvement.

$ curl -s https://netdoctor.e3d.ai/llms.txt

→ install steps, JSON schema, and the checkout API, all in one file the model can read directly.

Read llms.txt →

Free, or pay per report

The verdict is always free. Paying gets you an AI analyst's writeup.

500 credits = 1 paid-report run. Pay with a card below, or use e3d-netdoctor paid-report --card from the CLI directly — same checkout, no browser round-trip needed.

Free — template narrative

capture + report

$ e3d-netdoctor capture en0 30
$ e3d-netdoctor report ./capture.pcap

{
  "findings": {
    "verdict": {
      "headline": "Likely upstream/ISP",
      "confidence": "High",
      "rationale": "7 of 8 independent
        destinations affected, 3 providers"
    }
  },
  "narrative": {
    "source": "fallback",
    "executiveSummary": [
      "Top bandwidth conversation:
      10.0.0.5↔1.1.1.1 moved 42.1 MB
      across 3,204 packets."
    ]
  }
}

$5 — real AI narrative

paid-report --card

$ e3d-netdoctor paid-report --interface en0 --card

✓ payment confirmed — 500 credits
{
  "findings": {
    "verdict": {
      "headline": "Likely upstream/ISP",
      "confidence": "High",
      "rationale": "7 of 8 independent
        destinations affected, 3 providers"
    }
  },
  "narrative": {
    "source": "e3d-hosted",
    "executiveSummary": [
      "Your connection is almost certainly
      upstream — this pattern shows up
      across 3 unrelated providers, not
      just one flaky site."
    ]
  }
}

The verdict logic is identical either way — deterministic, rule-based, unit-tested, never "inconclusive." What paid-report adds is a real AI-written analysis of your findings (Claude or OpenAI, using your own key if you have one, e3d's otherwise), in one non-interactive call with a payment receipt attached — the free path always gets the plain-template writeup above, even if you happen to have an API key configured.

No card, no account

Capture & Report

runs on your machine

The full pipeline — capture, deterministic verdict, template-written report — runs free, forever, on your own hardware. paid-report adds a real AI-written analysis and a one-call, non-interactive flow.

Get started free →

Single Report

$5.00 / report