opensigma

Your Stripe data, in your own database. Free, open source, no Sigma bill.

A free, open-source replacement for Stripe Sigma and Stripe Data Pipeline. It mirrors your Stripe account into a Cloudflare D1 database so you can query it, join it against your own data, and build dashboards on top — without paying Stripe a second time for data you already gave them.

It replicates the core behaviour of stripe/sync-engine, but runs entirely on Cloudflare Workers + D1 + Queues instead of Postgres and a long-lived server.

What you get

• Every billing-core object as a queryable SQLite row — 25 synced Stripe resources, mirroring sync-engine's Postgres column set, ported to SQLite.
• Webhook-driven freshness. Rows update seconds after they change in Stripe. Out-of-order deliveries are dropped by a per-row last_event_at guard.
• Historical backload, paged from the Stripe REST API on an hourly cron, with a queue that retries anything webhooks missed.
• Plain SQLite. Query it with any SQLite client, any Worker, any BI tool. Join it against your own application tables.
• Read-only by design. The Stripe key it needs can't move money, refund anyone, or mutate your account — it can only read.
• Free-tier friendly. D1, Queues, and Workers cover most accounts at $0/mo.
• Ask questions in plain English with the bundled skill — and bring whichever model you like.

Ask in English, get SQL

This repo ships a Claude Code skill at .claude/skills/stripe-schema-query/. It reads the Drizzle schema, writes the SQL, runs it through wrangler, and renders the result.

/stripe-schema-query top 10 products by revenue last quarter
/stripe-schema-query MRR by month for the last year
/stripe-schema-query churn rate by plan, in-store vs online

The skill is just a prompt plus a workflow, so the model is your choice — Claude, GPT, Gemini, or something local. It works with any agent that supports skills.

opensigma vs Sigma vs Data Pipeline

	opensigma	Stripe Sigma	Stripe Data Pipeline
Price	$0 (Cloudflare free tier covers most accounts)	CA$14/mo (250 charges) → CA$621/mo (25k), then CA$0.028 per extra charge	CA$69/mo (1k charges) → CA$759/mo (25k), then CA$0.034 per extra charge — plus what you already pay for the warehouse
Where the data lives	Your D1 database (SQLite, you own it)	Stripe's servers — dashboard-only access	Snowflake / Redshift / Databricks / S3 / GCS / Azure (you provision and pay)
Query interface	Any SQLite client, any Worker, plus a natural-language skill that works with any LLM	Stripe's dashboard SQL editor + Stripe-managed AI	Whatever your warehouse supports
Join with your own app data	Yes — it's just SQLite, ship a Worker	No — Stripe data only	Yes, once it lands in your warehouse
Freshness	Seconds (webhook-driven)	Real-time (Stripe-side)	Batched, multi-hour latency
Build apps & dashboards on top	Yes — Workers bindings, any framework, any host	Limited — schedule reports, export, embed	Yes — whatever your warehouse supports
Vendor lock-in	None — MIT, runs on your Cloudflare account	Total — Stripe dashboard only	Partial — warehouse-portable but Stripe-gated
Source code	This repo, MIT-licensed	Closed	Closed
Cost at 100k charges/mo	$0	~CA$2,721/mo	~CA$3,309/mo + warehouse

Sigma and Data Pipeline figures are computed from the annual-billing tiers published on Stripe's Sigma pricing and Data Pipeline pricing pages (CAD, checked August 2026). Monthly billing costs more — Sigma's 250-charge tier is CA$21/mo rather than CA$14. Both charge overage per charge above the tier ceiling, which is what produces the curve above.

Quick start

pnpm install

# wrangler.jsonc is gitignored — start from the example
cp wrangler.jsonc.example wrangler.jsonc

# Create the D1 database, then paste the printed database_id into wrangler.jsonc
wrangler d1 create stripe_sync

# Create the queues
wrangler queues create stripe-sync-backload
wrangler queues create stripe-sync-dlq

# Set secrets
wrangler secret put STRIPE_API_KEY        # the read-only restricted key, see below
wrangler secret put STRIPE_WEBHOOK_SECRET # from the Stripe webhook endpoint

# Apply migrations
pnpm db:migrate

Then point your Stripe webhook at https://<your-worker-subdomain>/webhooks/stripe and subscribe to the event types listed in src/webhooks/dispatch.ts.

The Stripe restricted API key

Create a read-only restricted key at dashboard.stripe.com/apikeys/create?name=opensigma.

You want Read access on every resource, not just Billing Core — this project syncs charges, payouts, balance transactions, disputes, checkout sessions, payment methods, radar, and more. The fastest way:

1. Click Select all permissions at the top of the permissions list.
2. Switch the bulk selector from None → Read.
3. Scroll through and confirm everything is Read (leave Write unchecked).
4. Create the key and copy the rk_live_… (or rk_test_…) value.

A read-only key means a compromised Worker can't move money, refund anyone, or mutate your Stripe account — it can only read.

How the sync works

Two paths write to the same tables, as in the diagram above:

• Webhooks keep rows fresh in real time. Each row carries a last_event_at timestamp, so an out-of-order delivery is dropped rather than overwriting newer data.
• Cron + queue backfills history and re-heals anything webhooks missed. The hourly scheduled() handler enqueues a job per resource that isn't done; the consumer pages the Stripe REST API with the restricted key and upserts each page.
• D1 is the only datastore. Query it from any Worker, or via wrangler d1 execute.

To force a full re-backload:

pnpm backload:reset

The next cron tick re-fetches from the start. Webhook-updated rows stay protected by last_event_at.

The whole thing is one Worker exporting fetch, scheduled, and queue:

src/
  index.ts      # the Worker export — fetch / scheduled / queue
  app.ts        # Hono app; the webhook route is the only public path
  env.ts        # bindings and job types
  stripe.ts     # Stripe client factory
  webhooks/     # signature verification (handler) + event→upsert mapping (dispatch)
  events/       # incremental event polling, and the shared event processor
  backload/     # scheduled() enqueue, queue consumer, resource registry
  upserts/      # one file per Stripe resource, each guarded by last_event_at
  db/           # Drizzle client + schema.ts (the 28 tables)

What's in the database

28 tables in total — 25 synced Stripe resources plus 3 for bookkeeping.

Bookkeeping — stripe_events, backload_state, backload_parent_progress.

Synced Stripe resources — customers, products, prices, subscriptions, subscription_items, subscription_schedules, invoices, invoice_line_items, charges, balance_transactions, payment_intents, payment_methods, setup_intents, refunds, disputes, payouts, credit_notes, credit_note_line_items, checkout_sessions, checkout_session_line_items, coupons, promotion_codes, tax_ids, reviews, early_fraud_warnings.

Schema lives in src/db/schema.ts.

Querying your data

It's an ordinary SQLite database:

wrangler d1 execute stripe_sync --command \
  "SELECT COUNT(*) FROM invoices WHERE status = 'paid' AND created > strftime('%s', '2026-01-01')"

Or wire it into a Worker, a dashboard, or the BI tool of your choice. Nothing here phones home to Stripe for analytics — once the data is in D1, it's yours.

Development

pnpm db:migrate:local
pnpm dev          # local Worker
pnpm test         # vitest-pool-workers — real D1 and Queues bindings
pnpm typecheck

Migrations are generated, never hand-written: edit src/db/schema.ts, then run pnpm db:generate.

Built with

• Cloudflare Workers, D1, and Queues
• Hono for the webhook route
• Drizzle ORM + drizzle-kit for the schema and migrations
• stripe-node for signature verification and REST paging
• Vitest via @cloudflare/vitest-pool-workers — tests run in workerd against real D1 and Queue bindings

Schema and column sets follow stripe/sync-engine.

Contributing

Issues and pull requests are welcome — see CONTRIBUTING.md for setup, the checks CI runs, and the four touchpoints a new synced resource needs. By participating you agree to the Code of Conduct.

Found a security problem? Please follow SECURITY.md and report it privately rather than opening an issue.

Licence

MIT — see LICENSE.

---

_{Not affiliated with, endorsed by, or sponsored by Stripe. Stripe, Stripe Sigma, and Stripe Data Pipeline are trademarks of Stripe, Inc. Pricing figures are from Stripe's public pricing pages and may be out of date — check them yourself before making a decision.}