What is this?
Open-source metering and usage-based billing infrastructure: send usage events from your app, and Lago turns them into billable metrics, pricing, entitlements, invoices, and payments. Mistral AI, Groq, and PayPal use it publicly.
Open-source metering and usage-based billing infrastructure: send usage events from your app, and Lago turns them into billable metrics, pricing, entitlements, invoices, and payments. Mistral AI, Groq, and PayPal use it publicly.Is it a headless API, or does it have a UI too?Both: it's headless and API-first — REST API, SDKs, webhooks, and an MCP server for your code and agents — plus a web UI for finance, product, and ops teams when a human view is faster.
Both: it's headless and API-first — REST API, SDKs, webhooks, and an MCP server for your code and agents — plus a web UI for finance, product, and ops teams when a human view is faster.How do I get it running locally?Run the repo's Docker Compose demo — it needs Docker, curl, and jq (set LAGODEMOUIPORT/LAGODEMOAPIPORT if 8080 or 3001 are taken). For production, follow the self-hosted deployment guide or use the Helm charts.
Run the repo's Docker Compose demo — it needs Docker, curl, and jq (set LAGODEMOUIPORT/LAGODEMOAPIPORT if 8080 or 3001 are taken). For production, follow the self-hosted deployment guide or use the Helm charts.How does my application send usage events to it?Through the REST API, using official SDK clients for Node.js, Python, Ruby, and Go; webhooks push billing events back to your systems.
Through the REST API, using official SDK clients for Node.js, Python, Ruby, and Go; webhooks push billing events back to your systems.Does it charge my customers' cards itself?No — Lago deliberately keeps metering and pricing separate from money movement: it orchestrates payment providers like Stripe, Adyen, or GoCardless, including retries and dunning, so you still need a payment provider.
No — Lago deliberately keeps metering and pricing separate from money movement: it orchestrates payment providers like Stripe, Adyen, or GoCardless, including retries and dunning, so you still need a payment provider.Does a self-hosted instance send data back to Lago?Basic product analytics are collected by default, but no customer PII or financial data, and self-hosters can opt out.
Basic product analytics are collected by default, but no customer PII or financial data, and self-hosters can opt out.Any licensing gotchas if I self-host it or build on it?The platform is AGPLv3 — copyleft that matters if you offer it to others as a service. The Agent SDKs and MCP server live in separate repositories under MIT.
The platform is AGPLv3 — copyleft that matters if you offer it to others as a service. The Agent SDKs and MCP server live in separate repositories under MIT.What kinds of usage can it meter and bill?Anything you can emit as events: LLM tokens by model, input, output, cache, reasoning, or tool call; GPU, CPU, and storage consumption; API calls, transactions, seats, and custom events — plus prepaid credits with automatic top-ups and subscriptions with minimum commitments and overages.
Anything you can emit as events: LLM tokens by model, input, output, cache, reasoning, or tool call; GPU, CPU, and storage consumption; API calls, transactions, seats, and custom events — plus prepaid credits with automatic top-ups and subscriptions with minimum commitments and overages.What's the difference between Lago Direct and Lago Embedded?Direct is the standard mode: you meter and bill your own product's customers, like Mistral AI does. Embedded white-labels the same engine so platforms and marketplaces can offer billing inside their product under their own brand, like PayPal.
Direct is the standard mode: you meter and bill your own product's customers, like Mistral AI does. Embedded white-labels the same engine so platforms and marketplaces can offer billing inside their product under their own brand, like PayPal.Can AI agents work with Lago directly?Yes, it's agentic-first: an MCP server gives agents read/write tools for invoices, usage, customers, and payments, and Agent SDKs for Python and JS/TS instrument OpenAI, Anthropic, Mistral, and Gemini clients with under 5 ms p99 wrapper overhead.
Yes, it's agentic-first: an MCP server gives agents read/write tools for invoices, usage, customers, and payments, and Agent SDKs for Python and JS/TS instrument OpenAI, Anthropic, Mistral, and Gemini clients with under 5 ms p99 wrapper overhead.Is it production-ready for real scale?That's the design: idempotent ingestion deduplicated by transaction_id, atomic batches, isolatable async workers, Prometheus metrics, and SOC 2 Type II certification. Synthesia and Blacksmith run their billing on it publicly.
That's the design: idempotent ingestion deduplicated by transaction_id, atomic batches, isolatable async workers, Prometheus metrics, and SOC 2 Type II certification. Synthesia and Blacksmith run their billing on it publicly.What are similar projects?flexprice, QuantDinger, signoz, hyperswitch, invoiceninja, polar, langfuse, trueforge
flexprice, QuantDinger, signoz, hyperswitch, invoiceninja, polar, langfuse, trueforgeWhich topics is it in?pricing, subscriptions, events, billing, stripe, helm, clickhouse, payments, fintech, prometheus, ruby, agplv3
pricing, subscriptions, events, billing, stripe, helm, clickhouse, payments, fintech, prometheus, ruby, agplv3