See Every token, attributed.
Usage and estimated cost by department, team, project group and cost center. Observed usage, estimates and provider charges are kept apart.
AI agents you can see, afford and audit
Your AI agents are working
Know what they cost
Understand agent usage with Ork Meter, manage it across your organization with Ork Meter Enterprise, and coordinate engineering workflows with Orkestera. Built by senior engineers who help you put these systems to work.
See it · for developers · Launched October 23, 2026
Ork Meter
Ork Meter tracks your Claude Code and Codex usage in one live dashboard: tokens by day, by model and by project, valued at the vendors' published list prices and at the effective rate of the plan you actually pay for. Its plugin also trims noisy tool output before the agent reads it: on heavy-output tasks, that cut non-cached tokens 29.3% in Claude Code and 32.0% in Codex.
One dashboard, every agent
Claude Code and Codex report in through lightweight plugins. Usage metadata only; your conversations stay yours.
Clear numbers
Rate-card value of what you used, what your plan actually charges, and what the plugin trimmed, each shown with its assumptions and methods disclosed.
Per-model visibility
See which models your agents reach for and what each one costs, with list prices tracked nightly against vendor pricing pages.
Pricing
Metering is free for a limited time. Plugin optimizations are $20 per seat per month, free for your first 7 days. No card, no automatic charge. Savings shown on the dashboard are projections from our benchmarks; results vary and are not guaranteed.
Install in one line
Paste it into your terminal, then connect.
Run it in your own terminal. It installs the plugin for Claude Code and for Codex, whichever it finds. Needs Claude Code 2.1.280 or later; for Codex, the Codex CLI and Python 3.11, 3.12 or 3.13. Then run /ork-meter:connect in Claude Code; Codex pairs during the install.
Install the plugin, pair your workspace, and watch the first session land.
Go to orkmeter.com Install guideOwn it · for the organization
Ork MeterEnterprise
Know what your AI costs, and who’s spending it. One view of your organization’s AI token usage and estimated cost, by department, team and project, so finance, IT and engineering work from the same numbers.
Own Budgets every team can see.
Monthly token and dollar allowances for the organization and each unit, with advisory status at 80% and 100%. Nothing interrupts a developer’s session.
Prove A record you can hand to finance.
Owner, administrator, finance and viewer roles, a history of every administrative change, and CSV exports for chargeback.
A 30-minute walkthrough with a sample workspace. No setup on your side.
Book a demoRun it · the flagship
Orkestera
Agentic Workflow Management Suite
Our flagship platform for orchestrating teams of AI agents. Describe a goal in the Executive chat; Orkestera decomposes it into work and hands it to the Runner, which executes it as a workflow on the GraphAgent engine — reading files, running commands, calling models, and scoring its own results against continuous quality checks. Every run ends in a pull request on your repo. A human at the merge, not in the keystrokes.
Graph-based execution
GraphAgent runs each workflow as a directed graph of plugin nodes, not a linear script. Work moves through the graph as messages passed between nodes in both directions, and the engine executes and evaluates every node as the run proceeds. Graphs persist and re-run.
Composable plugin runtime
Every node is a plugin — a self-contained unit of behavior like reading a file, running a shell command, calling a model, or scoring a result. Extend what agents can do by adding plugins, not rewriting the core. Each node runs as its own isolated process; workflows persist and re-run from a terminal or the web.
Quality gates everywhere
Scoring lives in the graph — grading a result is just another node, so quality checks run inline, step by step. Every unit of work is evaluated before it lands, with a full audit trail and halt/resume controls.
Ends in a pull request
Never a silent push. The Runner opens a branch on your repo and the diff lands there for you to review. You're at the merge, not in the keystrokes.
Real-time observability
Watch streams every node transition, model call, and token — live on a Sprint board — so you always know what your agents are doing and what it costs.
Self-hosted & provider-agnostic
OAuth 2.1 throughout, with per-agent, least-privilege, short-lived credentials and MongoDB as your single data layer. Agents call any provider directly with your keys. Your models, your infrastructure, your data.
Orkestera is in active development with design partners.
Request early accessProof
Measured, not modeled.
−29.3%non-cached tokens in Claude Code1,524 heavy-output task attempts
−32.0%non-cached tokens in Codex5,369 heavy-output task attempts
≈ 0%on everyday coding (SWE-bench Lite)we publish the flat results too
Ork Meter plugin on versus off, 2026-10-22. Token counts, not a bill. Methodology, full results and limits.
What we do
Consulting, end to end
We wrote the playbooks and carried the pagers. Our practice spans the full delivery lifecycle — and, increasingly, the governance of agentic software systems.
Release Engineering & CI/CD
Trunk-based delivery, automated pipelines, and build systems tuned to your organization — so shipping is boring and frequent.
Cloud & Infrastructure
AWS architecture, infrastructure as code, and migration planning. Right-sized platforms without premature complexity.
Security & Compliance
Security built in from the first commit, the way auditors hope for and the law requires. Never bolted on after.
Observability & Reliability
Metrics, logs, traces, and dashboards that answer questions. SLOs your team can actually operate against.
Architecture & Delivery
Principal-level guidance for distributed systems and product delivery. Ship things that work — or don't ship them.
Agentic Engineering Enablement
Adopt AI coding agents with quality gates, audit trails, and human oversight. Governance that turns agent output into production software.
How we work
Rigor over speed
Ship things that work
Or don't ship them. Quality is a constraint, not a preference — every engagement leaves behind systems your team can operate without us.
Human near the loop
Automation should remove toil, not judgment. We design systems where people set direction and machines do the repetitive work — with oversight built in.
Security is table stakes
Encrypted in transit and at rest, least privilege by default, two-factor everywhere. The boring fundamentals, done every time.
No premature complexity
We add moving parts only when there's a clear, compelling reason. Simple systems fail less and cost less.
Who we are
A small, senior team
TH
Travis Hill
Chief Executive Officer
Twenty years leading engineering teams and running platforms at enterprise scale. Travis leads EDOS the way he's led every good team: with clear expectations, empowerment, and ownership.
JA
Jeremy Anderson
Chief Technology Officer
Twenty years building software for clients of every size, and an AWS Certified Solutions Architect. Jeremy works every layer — architecture, infrastructure, code — the bridge between intricate challenges and running systems.
DL
Dan Lambert
Chief Scientific Officer
Thirty years of engineering in unusual places: radar that sees through walls, NASA research programs, and augmented-reality games at his own studio, 256 Games. If it's hard and a little weird, Dan has probably built it.
Get in touch
Let's put your AI investment to work
Whether you need a delivery pipeline that stops hurting, a cloud migration done right, or a governed path to agentic engineering — leave your email and we'll reach out. You can also write us directly at info@edos.io.