Documentation

Everything FleetGrader does, it does offline.

This page mirrors the in-app Help screen line for line — when a behavior changes in the app, it changes here too. Nothing below is aspirational; every claim maps to a real screen, setting, or command.

Getting started

Your first report

Drop the export

Open New Report and drop a CSV or XLSX export from AppSpace or Korbyt — or a BrightSign log file. FleetGrader works out what it is; there's no format to pick by hand.

Get the report

Each run writes its own dated folder under Reports folder (set in Settings) with a report.docx, summary.xlsx, and, for supported types, a report.html.

Act on it

The report is a prioritized list of what needs attention and why, not a raw dump of the export you already had.

Tracking more than one site

Solo or Multi fleet memory

If you run reports for more than one client, switch Fleet memory to Multi in Settings (it defaults to Solo). Multi asks which client a log belongs to the first time it sees a new site, then remembers it — you're not re-identifying the same site on every run. Tick "Include this client's fleet history" on New Report to fold that history into the report you're generating.

Fleet Rollup

See every tracked site at once

Once at least one client is tracked, Fleet Rollup (sidebar, or the CLI for source checkouts) builds a single dashboard across all of them: pattern frequency, health tiers, and which sites a given issue has shown up on. It never shows one client's real hostname, serial, or display name to another — the dashboard is built from safe aggregate counts by construction, not a scrub pass run after the fact.

Drive it with an assistant (MCP)

Ask your AI assistant your fleet's status instead of opening the app

FleetGrader ships a local MCP server: four tools an AI assistant can call directly, over stdio, on your own machine. Nothing here is aspirational — every step below was run for real against the actual signed exe before this page was written. No source checkout, no Python install, no private repo access — the same FleetGrader.exe you already downloaded above does this too.

Set it up

Turn it on — it's off by default:
FleetGrader.exe --mcp enable
Register it with your MCP client, pointed straight at the exe wherever you put it. MCP is an open protocol — this works the same with any MCP-compatible assistant. Here's the command for Claude Code:
claude mcp add fleetgrader -- "C:\path\to\FleetGrader.exe" --mcp-server

On Mac, point at the binary inside the app bundle, not the .app itself — the app bundle doesn't respond to --mcp-server, the binary inside it does:

claude mcp add fleetgrader -- /Applications/FleetGrader.app/Contents/MacOS/FleetGrader --mcp-server
Confirm it connected:
claude mcp get fleetgrader

Should read Status: ✔ Connected. If it instead says the tools are disabled, step 1 didn't take — re-run it and re-check with FleetGrader.exe --mcp status.
Ask it something. In your assistant, ask in plain language — "what's my FleetGrader license status," "run a fleet health check," "how's client-001 doing." No slash commands, no flags to remember.

What the four tools actually do

Tool	Needs a license	Does
`license_status`	No	Current trial/license status. Calling it on a fresh install starts the trial clock — it's a real action, not a passive read.
`fleet_health`	Yes	Cross-client counts and health tiers (fire/soft/silent) aggregated over every tracked client.
`run_rollup`	Yes	Builds the real Fleet Rollup dashboard file to your reports folder and hands back its counts — never the file path. The file itself has every tracked client's real name in it; the assistant only ever sees safe totals.
`client_status`	No	One client's own counts, by ID. Never lists known IDs on a miss — an assistant can't fish for the rest of your client roster.

The part that isn't negotiable, verified

Every tracked client's real display name, hostname, and any domain alias never cross the wire to the assistant — the tools return safe aggregate counts by construction, not a scrub pass run after the fact. Verified for this page, not assumed: called client_status on a real tracked client with a real recorded identity, then grepped the raw JSON-RPC response for that client's real name and domain. Zero hits.

BrightSign log deciphering

What a raw BrightSign log actually means

A BrightSign unit's Diagnostic Web Server can hand you the raw log directly — it defaults off for security, so reading it locally means opening that up on the box. FleetGrader reads that same raw log and tells you what it means: whether a pattern is chronic-and-fine or genuinely new, without touching DWS and without the log leaving your network.

Licensing

Trial, then a key

New installs get a 14-day trial, no key required. After that, activate a license key (starts with FGRADE-) in Settings, or buy one for $99/yr flat — one price, no per-device metering, unlimited reports — from the button next to it, or from the pricing section on the homepage.

Get help

Built and supported by one person — me.

Email with a question, a bug, or a fleet export that trips it up. Real replies, not a ticket queue.

[email protected]