About Meter for Drivers & Fleets
The Meter plugin gives Claude Code, Codex, and other agents access to Meter taxi booking and fleet management through a hosted MCP server. It combines six skills with account-role-based data access and previews and confirmations for changes.
Key features
- 01Review fleet trips and earnings
- 02Book, track, and cancel rides
- 03View driver trips and earnings
- 04Guide fleet setup
- 05Preview and confirm settings changes
- 06Explain API and data fields
Use cases
- 01Summarize fleet earnings and trips
- 02Confirm price and pickup details before booking
- 03Change fleet or personal settings
- 04Review driver earnings and pay stubs
MCP support
The hosted MCP server provides Meter tools to clients after OAuth 2.1 sign-in, covering rides, fleets, drivers, and settings.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://www.meterapp.co/api/mcp/fleetclaude mcp add --transport http co-meterapp-meter https://www.meterapp.co/api/mcp/fleet{"mcpServers":{"co-meterapp-meter":{"type":"http","url":"https://www.meterapp.co/api/mcp/fleet"}}}Open in Cursor{"servers":{"co-meterapp-meter":{"type":"http","url":"https://www.meterapp.co/api/mcp/fleet"}}}Open in VS Code{"mcpServers":{"co-meterapp-meter":{"type":"http","url":"https://www.meterapp.co/api/mcp/fleet"}}}{"mcpServers":{"co-meterapp-meter":{"serverUrl":"https://www.meterapp.co/api/mcp/fleet"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEMeter — plugin for Claude Code and Codex
Meter is the taxi meter and fleet platform: drivers run a meter in the Meter app, fleets manage drivers, vehicles, fares, scheduled bookings and money, and riders book rides from fleets. This plugin gives your agent the view you already have in the Meter apps, through the hosted MCP server, with your own Meter sign-in. One repository, native packaging for each client, plus the vendor-neutral Open Plugins manifest.
No API keys. The first time a tool is used, your agent opens a Meter sign-in in the browser (OAuth 2.1 with PKCE). The token it receives is exactly as powerful as you are: owners and admins see fleet-wide data, drivers and riders see their own.
Install
Claude Code
/plugin marketplace add MeterApp/meter-plugins
/plugin install meter@meterapp-meter
Codex
codex plugin marketplace add MeterApp/meter-plugins
codex plugin add meter@meterapp-meter
Skills only, any agent
npx skills add MeterApp/meter-plugins
Server only, no skills
claude mcp add --transport http meter https://www.meterapp.co/api/mcp
Start a new session, then try:
"How much did my fleet earn last week, and which driver had the most trips?"
"Book me a ride to the airport at 6am with my usual taxi company."
Claude and ChatGPT (no terminal)
Add a custom connector with the URL https://www.meterapp.co/api/mcp in the app's connector settings and sign in to Meter when asked. The skills are optional there: the server explains its own tools and asks for a preview and your yes before anything changes.
Keep it up to date
- Claude Code updates the plugin by itself at startup once auto-update is on for its marketplace, which Claude Code leaves off for marketplaces outside Anthropic's: open
/plugin, go to Marketplaces, choosemeterapp-meterand select Enable auto-update. By hand:/plugin marketplace update meterapp-meter, then/plugin update meter@meterapp-meter, then restart (or/reload-plugins). - Codex:
codex plugin marketplace upgrade meterapp-meter. - Skills only:
npx skills update.
Added this marketplace before 1.4.0, when it was called meterapp? CHANGELOG.md shows how to move your install to the new name.
The CLI, @meterapp/meter, installs its own new versions in the background, and meter update updates it now and refreshes this plugin wherever Claude Code or Codex has it installed.
What you get
Six skills and the hosted MCP server.
| Skill | Job |
|---|---|
meter-ride | Booking, following and cancelling your own taxi ride: confirmed addresses, the price first, your yes before booking |
meter-fleet-insights | Questions about a fleet's trips, bookings, drivers, vehicles, earnings and accounting |
meter-fleet-onboarding | Walking a new fleet owner through setup with the onboarding checklist and settings tools |
meter-settings-changes | Changing fleet, booking, ride-option or personal settings: dry run, confirm, apply |
meter-driver-personal | An individual driver's own trips, earnings, pay stubs, availability and profile, and a rider's past rides |
meter-api-explainer | "How does this work / what does this field mean", and drafting integrations |
The MCP server at https://www.meterapp.co/api/mcp provides the tools listed in references/tools.md; the objects they return are in references/data-model.md. Both files are generated from the server's own catalog. The server also describes itself: describe_api returns a domain index by default; pass domain to browse its tools or tool for full detail. explain_data explains any result field.
Authentication
| Surface | Credential |
|---|---|
| Hosted MCP | OAuth 2.1 access token obtained by the host from your Meter sign-in |
| CLI | npx @meterapp/meter login (same sign-in, stored locally); meter call <tool> --dry-run previews any write |
| Browsers, config files, prompts | Never a token |
The host discovers the sign-in from the server's 401 challenge and the metadata at https://www.meterapp.co/.well-known/oauth-protected-resource. Nothing in this repository contains or expects a credential. Details in references/authorization.md.
Safety model
- Read first, ask second.
whoamitells the agent who is signed in and which fleets and roles they have; ids come from results, never from guesses. - Every settings change is a dry run first.
dry_rundefaults to false, so the agent passesdry_run: true, shows the exactchanges(field,from,to), waits for an explicit yes, then repeats the call to apply. Fares are never changed without stating the from and to values. - A permission error is final for that person. The agent switches to a tool that fits their role instead of retrying.
- Results never contain payment-provider ids, GPS traces, internal notes or other people's contact details beyond what the role allows. See references/safety.md.
- Rides are booked only for the signed-in rider, only after a preview of the price, pickup time, payment and cancellation policy, and only on an explicit yes. Nothing is paid in the chat: when a ride needs paying for, the agent gives you a Meter checkout link (valid about 30 minutes, only for you) and the ride is booked once you have paid on meterapp.co; it then checks the booking before saying so. Rides the company takes payment for after the trip, and free rides, are booked in the chat. The agent never asks for or types card details. Cancelling shows the fee first.
What the plugin sends, and where
Everything goes to one place: the hosted Meter MCP server at https://www.meterapp.co/api/mcp, authenticated by your host with your own Meter sign-in. The agent sends the arguments of the tools it calls, which can include addresses you ask it to search or book, ride times and booking ids. Nothing in this plugin reads files, environment variables or credentials from your machine, and nothing is sent anywhere else. How Meter handles that data: privacy policy. To disconnect, remove the connector in your host or revoke it in Meter under Account, Connected apps.
Development
python3 scripts/validate_repo.py
claude plugin validate . --strict
codex plugin marketplace add . --json
This repository is published from the Meter source repository. Please open an issue here, or write to [email protected], rather than sending a pull request; changes are overwritten on the next release.
Links
Docs · MCP · Plugins · Book rides · Privacy and terms · MCP Registry entry · agents.md · llms.txt · openapi-mcp.json · [email protected]
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.