Connect the AI you pay for to the work it helps.
Two sets of connectors and one join between them. Every connector reads amounts, people, dates and states. None of them reads a prompt, a file or a ticket body.
Implementation and public availability are different. Implemented means the capability is in the product. Built, not yet live means the code is written and tested, but has not yet run against a real account or is not switched on in our hosted service. Planned means intended coverage that is not built. Contact us to confirm availability for your environment.
Assistants, APIs and agents.
Billing and usage from each vendor, settled into one record and resolved to the people who ran it.
- Built, not yet live
Anthropic
Daily API cost and currency. A breakdown by model is coming.
- Built, not yet live
OpenAI
Daily API cost for the organization. A breakdown by project, line item and key is coming.
- Built, not yet live
Cursor
Team spend per user through the Admin API.
- Built, not yet live
GitHub Copilot
Usage, billed cost and list price from GitHub's billing usage report.
- Planned
ChatGPT Enterprise
Seats and workspace usage.
- Planned
Claude for Work
Seats and workspace usage.
- Planned
Gemini and Google Workspace
Seats and usage across Workspace.
- Planned
Microsoft 365 Copilot
Seats and usage by person.
- Planned
Windsurf and Amazon Q
Seats and usage for more coding assistants.
- Planned
AWS Bedrock, Azure OpenAI, Vertex AI
Model spend on your cloud accounts.
- Planned
Your own billing exports
Any spend no API reaches, filed in the same record. Upload is planned.
- Implemented
The desktop client
Claude Code, Codex and Cursor sessions and plan limits, read on each machine. Public downloads are coming soon.
The record of what your people finished.
Work items with a person and a date on them, joined to the same people in the same period as the spend.
- Built, not yet live
GitHub
Pull requests, commits, reviews and changed lines.
- Built, not yet live
Jira
Issues, epics, projects and resolution.
- Built, not yet live
GitLab
Merged merge requests, who wrote them and how long each took to merge.
- Built, not yet live
Linear
Issues with state, priority, assignee and how often one reopened.
- Planned
Azure DevOps
Work items, pull requests and boards.
- Built, not yet live
ServiceNow
Incidents, changes, requests and tasks, with priority and time to resolve.
- Built, not yet live
Salesforce
Opportunities and cases, with stage, amount and owner.
- Planned
HubSpot
Deals, tickets and activity.
- Planned
Databricks jobs
Jobs and workloads beside the model spend that ran them.
- Planned
Trello
Personal boards and completed tasks as work context.
- Planned
Slack work signals
Declared collaboration and workflow metadata as team context, not messages treated as completed work.
Who is on which team, and how they sign in.
Spend and work meet through identities you declare, so the join is never a guess.
- Built, not yet live
Google sign-in
One click for people and small teams. Sign-in is not open yet.
- Built, not yet live
Microsoft sign-in
Work accounts with verified domains. Sign-in is not open yet.
- Built, not yet live
Email link
No password to keep. Sign-in is not open yet.
- Built, not yet live
Okta
Enterprise sign-in, and groups mapped to teams.
- Planned
Microsoft Entra ID
Enterprise sign-in and groups.
- Planned
SCIM provisioning
Teams kept in step with your directory.
Run it where your data already is.
The same record in every profile.
- Built, not yet live
Hosted by ai-cur
Runs on our servers; sign-in is not open yet. Contact us for hosted availability.
- Implemented
Your own servers
One docker compose deployable inside your network.
- Implemented
Databricks App
Delta tables governed by Unity Catalog, with row filters per role.
- Implemented
ServiceNow scoped app
The cost record, ingestion and role-based views inside your instance.
- Implemented
Finance export
One chargeback CSV per month and currency that ties out to the page. Six of its seven columns use FOCUS 1.4 names; the seventh is our cost centre.
- Planned
Snowflake and BigQuery
The record landed in your warehouse.
Act on it in the tools you already run.
ai-cur stays out of the request path, reading billing through each vendor's reporting API and never handling a model request.
- Implemented
Cap decision endpoint
Any gateway or key issuer reads the decision and enforces it.
- Planned
LiteLLM
Spend and keys read directly, caps pushed back.
- Planned
Slack and Microsoft Teams
Budget, cap and idle-seat alerts where your team talks.
- Built, not yet live
Email alerts
Threshold and scheduler notices. On your own server, one email destination for the whole deployment.
- Built, not yet live
Webhook
On your own server, one webhook for the whole deployment receives every alert in place of email. Destinations per team or account are planned.
Missing the system your teams work in?
Tell us which one. The join needs a work item with a person and a date on it, and nothing else.
