Halton Meter Docs

Know exactly what your
AI tools are spending.

Halton Meter is a local proxy that meters every LLM request (tokens, model, and real cost) straight to a SQLite file on your machine. Set it up in a minute.

Install in 60s First request halton-meter report Troubleshooting
Get started How it works
4
Providers metered
~60s
To first capture
0
Bytes leave your box

Meter every token. Locally.

The documentation for Halton Meter, the proxy that turns invisible LLM spend into a cost report you actually read. Everything runs on your machine.

Get started Install guide
v0.4.1 · public beta 4 providers metered 0 bytes leave your box
01 Quickstart

From zero to a cost report in 60 seconds

Full guide →
1

Install the package

uvx runs the latest wheel with zero Python prerequisite. macOS is stable; Linux and Windows are in public beta.

$ uvx halton-meter init --apps
$ uv tool install halton-meter
$ pipx install halton-meter
2

Initialise your meter

Pick how much to capture. --apps covers your IDEs without touching the browser. Start there if unsure.

init 
$ halton-meter init --apps
  1/8 Generating mitmproxy CA certificate…   ✓
  2/8 Trusting cert in Keychain…             ✓
  4/8 Installing launchd supervisor…         ✓
  7/8 Daemon healthy  127.0.0.1:8081          ✓
✓ init complete   mode: apps
3

Read your first report

Open a new terminal, run any prompt through your tools, then pull the numbers.

report 
$ halton-meter report --today
02 Browse

Everything, by topic

Getting started
3 pages

Install, send your first metered request, and configure projects.

InstallFirst requestConfigure projects
Concepts
6 pages

How the proxy works, project tagging, the SQLite schema, and pricing.

Proxy modelProject taggingSQLite schema
CLI reference
9 pages

Every command, flag, and exit code for the halton-meter binary.

initstartstop
Providers
4 pages

Per-provider coverage: Anthropic, OpenAI, Gemini, and Grok.

AnthropicOpenAIGemini
Operations
5 pages

Day-two running: logs, TLS trust, upgrades, and troubleshooting.

TroubleshootingHow clients see your CAUpgrade
Security & privacy
2 pages

What stays local, the threat model, and how cert trust is scoped.

Local-only guaranteeDisclosure
Halton Meter Cloud
10 pages
Cloud

Optional sync for 90-day trends, team views, and cross-machine rollups.

OverviewThe cloud dashboardConnect your meter
03 Providers

Every LLM request logged, out of the box

Pricing methodology →
Anthropic Claude · live pricing OpenAI GPT · live pricing Gemini Google · live pricing Grok xAI · live pricing
04 Start here

Hand-picked pages

Curated
01 Install Halton MeterGetting started 02 halton-meter reportCLI reference 03 Choosing an install modeGetting started 04 Pricing methodologyConcepts 05 TroubleshootingOperations
What's new

Changelog

All →
0.4.1 Adds pricing for Cursor Composer 2Jun 8, 2026 0.4.0 First-class Linux and Windows support, plus cross-platform safety fixesJun 8, 2026 0.3.11 Cloud body uploader policy-rejection handling, new pricing/body status commands, and a covering index for fast body upload statusJun 7, 2026
05 Stuck?

We treat docs gaps as bugs

Troubleshooting

Cert errors, daemon loops, missing captures: every known failure with a fix.

Open guide →

GitHub

The bundled dashboard is open source (Apache 2.0). Read the source, file an issue, send a PR.

haltonlabs/halton-meter-dashboard →

Email the maintainer

Stuck for more than five minutes? That's a docs bug. We fix the page, not just your question.

operator@haltonlabs.com →