Know your Copilot burn before GitHub does.

BurnRate pulls your daily Copilot usage from GitHub, keeps the raw JSON for audit history, and forecasts where your budget will land at month end. It watches. It does not enforce.

Quick start View on GitHub

Raw payloads are stored before anything is parsed
Runs on PostgreSQL in production or SQLite locally
Ask questions through Copilot Agent Skills

Sample run

$ @burnrate /forecast
✓ loaded 28 days of usage data
✓ raw reports preserved in raw_reports
→ projected month-end burn: 78%
→ risk: within budget

For engineers watching Copilot spend.

Reports, forecasts, and audit history. Nothing more.

Raw-first storage

Keep the source payload before parsing. If the schema changes, your history stays intact.

Raw JSON is kept exactly as GitHub sent it
Parsing happens separately from storage

Capabilities

The core jobs BurnRate does with your local data.

Forecasts: project month-end burn from actual usage
Budget alerts: get notified before you hit limits, without writing to GitHub
Dual databases: PostgreSQL in production, SQLite locally, chosen from DATABASE_URL

Copilot Agent Skills

Run forecasts, classify users, or check budget sync from chat using plain language.

Chat commands on demand
They call the same CLI scripts you use

Install, configure, run.

These are the same commands the CLI uses. Pick Postgres or SQLite.

Install

$ git clone https://github.com/mhenke/BurnRate.git
$ cd BurnRate
$ npm install

Configure

DATABASE_URL=postgresql://user:pass@localhost:5432/burnrate
GITHUB_TOKEN=ghp_xxx
GITHUB_ENTERPRISE=my-company
GITHUB_ORG=my-org

Create a GitHub Personal Access Token (classic) with read:org permissions to fetch Copilot reports.

Run

# Run migrations to set up database tables
$ npm run migrate
# Ingest Copilot usage reports from GitHub API
$ npm run ingest
# Generate burn forecasts and budget projections
$ npm run forecast

From GitHub API to forecast.

History is stored raw, then parsed and aggregated in separate steps.

1GitHub APIDaily Copilot usage reports
2ETLRaw storage, parsing, normalization
3DatabasePostgreSQL or SQLite tables
4ForecastUsage trend and month-end burn
5AlertsBudget threshold notifications

Few env vars, one config file.

The same settings the runtime actually reads.

Environment variables

Environment variables for configuring BurnRate
Key	Purpose
`DATABASE_URL`	PostgreSQL or SQLite connection string
`GITHUB_TOKEN`	GitHub PAT with `read:org` scope
`GITHUB_ENTERPRISE`	Target enterprise slug
`GITHUB_ORG`	Target organization slug
`SLACK_WEBHOOK_URL`	Optional budget alert destination

Value tiers

teams:
  - name: Platform Engineering
    value_tier: extreme
  - name: Data Science
    value_tier: high
  - name: Internal Tools
    value_tier: medium

Observe-only, local-first security

Your GITHUB_TOKEN and Copilot usage data stay inside your environment. They go to your self-hosted database, not to us. BurnRate does not send tokens or raw logs anywhere else, and it never writes back to GitHub.

Small CLI, small chat surface.

Available BurnRate CLI commands and their corresponding Copilot Agent Skills
CLI Command	Copilot Agent Skill	Description
`npm run ingest`	`@burnrate /etl`	Fetch and store today's Copilot reports
`npm run forecast`	`@burnrate /forecast`	Generate burn forecasts from stored data
`npm run classify`	`@burnrate /classify`	Classify users by consumption/value tiers
`npm run budget-sync`	`@burnrate /budget-sync`	Sync budget limits and send alerts
`npm test`	None	Run the test suite