Know exactly where your AI budget is going
Most AI spend trackers stop at the invoice. Invoices sit downstream of activity. They don't show you why, or whether it was worth it. Flowstate connects every token to a person or service, a project, a cost center and an outcome. Across developer AI and production AI.
Two faces of AI spend, in one model
Most teams measure neither well. Flowstate covers both, because the economic questions are the same.
Developer AI
Claude Code, Cursor, Copilot, Windsurf
Interactive, episodic, attributable to a person. Usually billed per seat plus overage. Tied to projects, sprints and tickets.
Production AI
Chatbots, document extractors, video pipelines
Autonomous, high-volume, attributable to a service. Usually billed per token. Tied to product lines, customer actions and feature contribution margin.
Session classification, done for you
Every AI coding session is analyzed and classified automatically. The classifier reads prompts, file paths, git context and tool usage to categorize the work and match it to the right project, so spend flows through to the correct cost center.
The classifier categorizes the work
The classifier works out what was done and matches it to a project via ticket IDs, branch names and file paths.
The project's cost center sets CapEx or OpEx
Once a session is attributed to a project, spend flows through to the project's cost center. You set CapEx or OpEx on the project. Flowstate handles the rest.
AI spend is growing, but nobody knows where it's going
The trigger is familiar. You moved to an enterprise AI plan and the bill went up for the same headcount. Finance sees a growing "AI tools" line with no way to attribute it to teams or projects. Engineering leaders can't answer basic questions. Which teams are driving the Copilot bill? Is our Claude API spend producing anything? Should we invest more in agents or more in people?
Without attribution, AI spend is a black box, and black boxes get cut first when budgets tighten.
Full visibility into AI spend
Provider-level breakdown
See spend by provider: GitHub Copilot, Anthropic, OpenAI, custom agents. Know exactly where each dollar goes.
Project attribution
AI spend is attributed to projects using ticket IDs, branch names and file paths. Drill into any project to see per-session cost, category and confidence.
Trend analysis
Track AI spend over time with date range selectors. Spot runaway costs before they become quarterly surprises.
Token-level detail
See input and output token usage per agent, per project. Understand what you're spending, and why.
Full picture from both ends
Flowstate pulls cost and usage data directly from AI providers, and captures session-level telemetry from developer machines. What's being spent, and where it's going.
Provider API integrations
We pull billing and usage data directly from your AI providers. That's the cost side of the equation, with nothing to install.
Session telemetry CLI
Our open-source CLI identifies and configures telemetry on developer machines, whatever AI tooling they use. It captures session-level data for classification.
MDM-ready rollout
The Flowstate CLI identifies and configures telemetry on developer machines, whatever AI tooling they choose. Roll it out across your org with your existing MDM tooling: Jamf, Intune, Kandji, or plain shell scripts.
- ✓ Silent install via Homebrew or direct binary
- ✓ Auto-detects installed AI tools and configures telemetry hooks
- ✓ Zero-config for developers, no manual setup required
- ✓ Open source: audit the code, contribute, or fork it
Stop flying blind on AI spend
Book a demo and see exactly where your AI budget is going, and whether it's worth it.