§ 01 Three lanes
Daemon, Cloud, Enterprise.
Daemon is the local-binary — free forever, runs on your machine. Cloud is the hosted tier at cloud.haltonmeter.com that rolls up multiple Daemons for a team. Enterprise is self-hosted Cloud for data-residency workloads.
Free. Forever.
A local binary that runs on your machine. Captures every LLM call, attributes it to a project, costs it against published prices. The bundled dashboard is open source under Apache 2.0.
- − 6 adapters across 4 providers (Anthropic, OpenAI, Gemini, xAI)
- − OAuth surfaces — ChatGPT, Gemini Code Assist
- − Smart Attribution v0.3 (git repo · container · sandbox bundle)
- − Local SQLite at ~/.halton-meter/db.sqlite
- − Loopback HTTP API on 127.0.0.1
- − CSV / JSON export from the local API
- − Dashboard at localhost:3000 (Apache 2.0)
- ○ Multi-machine local rollup (your laptops, no cloud)
- ○ Deeper provider coverage (Mistral, Cohere, Groq)
- ○ Self-hosted aggregation for the privacy-first crowd
− shipping · ○ planned
Solo $16/mo · Team from $36/mo · Business+ by enquiry
Hosted dashboards that roll up every developer's local daemon into one view. Reconciliation against provider invoices. The audit log every client procurement team eventually asks for. Live at cloud.haltonmeter.com.
- − Cross-machine team rollup (one hosted view)
- − Reconciliation: captured spend ↔ provider invoices
- − Audit log of every captured request
- − Continuous optimisation recommendations
- − Historical trending — 3-month, 6-month, YoY
- − Branded PDF cost reports for clients
- − Per-project, per-developer rollups
- − Role-based view (founder / lead / IC)
- ○ S3 / GCS scheduled export (Team plan)
- ○ Webhook alerts on budget thresholds
− shipping · ○ planned
By contract.
Self-hosted Cloud for workloads that cannot leave the building. Your VPC, your keys, your Postgres — no traffic routed through Halton Labs infrastructure. Deploy via Helm chart or Terraform module.
- − Your VPC, your keys — no outbound data
- − SAML 2.0 + SCIM provisioning (Okta, Azure AD, JumpCloud)
- − Append-only audit log — every login, export, policy change
- − Custom data retention — your retention window
- − 99.9% uptime SLA on hosted components
- − Dedicated Slack channel + direct engineering access
- − Annual security review with findings report
- − DPA + SIEM export
- ○ SOC 2 Type II (on customer procurement request)
− shipping · ○ planned
§ 02 What we will not build
The anti-list.
The build plan is honest only if the no-list is honest.
These are commitments, not omissions: telling you what you are buying into.
- ×
A marketplace
No third-party adapters store, no plugin economy. Adapters live in the daemon, vetted by us.
- ×
Agent orchestration
Halton Meter measures spend. It is not a runtime, a router, or a control plane.
- ×
A generic observability platform
No traces, no logs, no APM. LLM cost is the surface; the page does not creep.
- ×
Vendor-cloud capture of your request bodies
The Solo daemon stores bodies locally. The Team tier rolls up metadata and computed cost — bodies stay on your machine unless you opt them in.
- ×
A mobile app
You do not need a phone to read a cost report. The dashboard is responsive; that is the contract.
- ×
A free hosted tier of the cloud
The daemon is free forever and genuinely useful on its own. The cloud tier covers a real cost; we will not subsidise it with a freemium funnel.
§ 03 Provenance
If we change the plan, the date moves.
Daemon and Cloud are shipping today. Cloud went live on 22 May 2026 — a quarter ahead of the original Q3 2026 commitment.
Capabilities marked “planned” ship when they ship; we will not pre-announce them.
Enterprise shipped alongside the cloud tier. If your workload has data-residency requirements the public cloud tier cannot meet, talk to us.
Start at cloud.haltonmeter.com.
Questions about a specific capability, or Enterprise:
operator@haltonlabs.com.
