Lago

Agentic-first monetization infrastructure for AI products
Meter tokens, compute, API calls, or any product usage.
Turn usage into pricing, credits, entitlements, invoices, payments, and revenue.
Use Lago Direct for your own product, or Lago Embedded to power monetization for your customers.

Talk to us about Lago Cloud · Build with your AI agent · Run the AI billing demo · Documentation

“Lago has been able to follow the pace of our releases and has allowed us to focus on what we do best.”
Timothée Lacroix, CTO at Mistral AI · Read the customer story

Build with your AI agent

Give your coding agent Lago's public documentation, then choose the API, CLI, or MCP for your workflow. Paste this prompt to start:

Read https://getlago.com/llms.txt and follow the relevant Lago guides for this
project. Use https://swagger.getlago.com/openapi.yaml for exact API fields.
Help me design and implement a billing integration. Ask before live access.

These interfaces work independently. Application code uses an SDK or the REST API; no Lago plugin is required. Confirm the target organization and environment, keep API keys out of prompts and source control, and review financial changes before execution.

Follow the agent setup guide for connection details and validation, or explore the website overview. To see real usage become a charge locally, run the maintained demo below.

See Lago price an AI workload

Run the maintained demo from this repository:

./examples/agentic-ai-demo/run.sh

It starts the Lago version that matches this checkout, creates a disposable local organization, and prices three illustrative AI requests. Each request sends one input-token event and one output-token event:

3 AI requests
5,000 input tokens  x $0.000002 = $0.01
1,250 output tokens x $0.000008 = $0.01
Lago usage total                   = $0.02

Docker Compose starts an isolated Lago service, then a small script seeds and verifies the example through Lago's API. It retrieves current usage, independently reconciles the result, and retries one transaction to confirm that usage does not increase. Everything stays on your machine; the demo does not access Lago Cloud. The bundled credentials and API key are disposable and only intended for this local demo.

Log into http://localhost:8080 with [email protected] / agentic-ai-demo-local-password, then open Customers → Agentic AI Demo Customer → Agentic AI Demo subscription → Usage.

Requirements: Docker, curl, and jq. If ports 8080 or 3001 are occupied, set LAGO_DEMO_UI_PORT and LAGO_DEMO_API_PORT before running. The demo keeps Lago running so you can inspect the customer, metrics, plan, subscription, events, and usage in the UI. Remove its isolated Compose project and data volume when you are done:

./examples/agentic-ai-demo/run.sh --cleanup

Run Lago's maintained Agentic AI demo by following this README. Keep it local;
do not modify source files or unrelated Docker resources. When it passes, give
me the UI URL, token usage and charges, idempotency evidence, and cleanup
command. Then offer to connect this agent to Lago's local MCP server.

What is Lago?

Lago is the programmable system between product usage and revenue. Send events from your application, turn them into billable metrics, apply pricing and entitlements, then generate invoices and collect payments.

Usage events -> Metering -> Pricing and credits -> Entitlements -> Invoices -> Payments -> Revenue

Use Lago to launch and change pricing without rebuilding billing:

• tokens by model, input, output, cache, reasoning, or tool call
• GPU, CPU, storage, and other compute consumption
• API calls, transactions, seats, active users, or custom events
• prepaid credits with automatic top-ups
• subscriptions with allowances, minimum commitments, and overages
• self-serve plans and negotiated enterprise contracts in the same system

Lago is headless and API-first. Your application, internal tools, and agents use the same billing primitives through the REST API, SDKs, webhooks, and MCP server. Product, finance, and operations teams can use the Lago UI when a human interface is faster.

How Lago fits into your stack

Lago keeps metering and pricing independent from payment processing. Connect Stripe, Adyen, GoCardless, or another provider without making its product catalog your source of truth.

Lago Direct and Lago Embedded

The same Lago engine supports two operating models:

Lago Direct is the standard way to use Lago: send your own product usage, model your pricing, and bill your customers. Lago Embedded uses the same primitives to let platforms, marketplaces, AI builders, fintech products, and developer tools offer metering and billing to their customers.

Both models start with Lago's open-source billing engine and can be paired with Premium deployment, customization, security, and support options. With Lago Embedded, you control the customer experience and which capabilities your users can access.

Agentic-first by design

Agentic-first means Lago's billing model is available as structured, inspectable interfaces instead of being trapped in a dashboard.

The MCP server exposes read and write tools and inherits the permissions of its Lago API key. Treat agent credentials as privileged, control who can access the agent, review tool calls, and keep human confirmation around sensitive billing changes.

Platform

Explore metering, billing and invoicing, entitlements, cash collection, revenue analytics, and integrations.

Production characteristics

• Idempotent ingestion. Lago deduplicates usage events by transaction_id, so retrying an event does not bill it twice. Read the ingestion guide.
• Atomic batches. If one event in a batch is invalid, Lago rejects the batch and persists none of its events. Read the batch API reference.
• Asynchronous processing. Dedicated workers can isolate events, billing, payments, invoices, webhooks, PDFs, alerts, and analytics as volume grows. Review the architecture.
• Observable operation. Lago exposes Prometheus metrics for APIs, queues, workers, events, billing, webhooks, and dependencies. Review monitoring.
• Deployment control. Lago Cloud runs and updates the platform for you. When you need to run it yourself, self-host Lago in your cloud or deploy it on-premise; self-hosted production deployments can separate and scale stateful services and workers independently.
• Security. Lago is SOC 2 Type II certified and supports self-hosting for teams that need infrastructure and data control. Review security.

Chosen by teams operating at scale

Mistral AI, Groq, PayPal, Synthesia, Blacksmith, and other teams have publicly selected Lago for billing. The figures below describe the companies using Lago, not volume processed by Lago.

Sources for company context: Mistral AI, Groq, Blacksmith, Synthesia, and PayPal.

Deployment options

Lago Cloud and self-hosted Lago differ in deployment, not features. Your code uses the same API and SDKs either way; only the base URL changes.

1. Lago Cloud (recommended)

With Lago Cloud, there is nothing to deploy or maintain. Lago runs and updates the platform for you in US or EU regions and provides commercial support, so your team never operates billing infrastructure. Talk to us about Lago Cloud.

2. Self-hosted and on-premise

Run the open-source platform from this repository (AGPLv3) in your cloud or on your own infrastructure when you need to inspect, extend, and operate billing yourself. Start with Deploy Lago below and the self-hosted deployment guide.

Also available

Deploy Lago

To self-host Lago or deploy it on-premise, start with a source-based local deployment:

git clone --depth 1 https://github.com/getlago/lago.git
cd lago

echo "LAGO_RSA_PRIVATE_KEY=\"$(openssl genrsa 2048 | openssl base64 -A)\"" >> .env
docker compose up -d

[!IMPORTANT] Upgrading the front image. From v1.56.0 (getlago/lago-front#4359), the front container runs nginx as uid 65532 (not root), with a read-only root filesystem, and listens on 8080 (8443 for TLS) instead of 80 (443). The docker-compose.yml in this repository already maps the host ports to the new container ports. If you run your own Compose file, Kubernetes manifests or reverse proxy:

• Point it at container port 8080. The host port can stay 80.
• Give /tmp a writable mount (a tmpfs or an emptyDir). The container writes its pid file, env-config.js and csp.conf there.
• If you mount your own nginx default.conf, listen on 8080/8443, and copy the env-config.js location, the dotfile location (it keeps /.env.sh private) and the include /tmp/csp.conf lines from extra/nginx-selfsigned.conf. Do not set ssl_dhparam: the image cannot load DH parameter files.
• Make TLS keys readable by uid 65532 (extra/init-selfsigned.sh and extra/init-letsencrypt.sh do this).
• Remove any command: override for the front service. It never reached nginx; use the entrypoint: in docker-compose.yml instead.

Use the self-hosted deployment guide for persistent storage, SMTP, TLS, object storage, Redis, database configuration, and upgrades. For larger deployments, use the Lago Helm charts and configure dedicated workers.

SDKs and developer resources

• Getting started
• API reference
• OpenAPI specification
• Pricing templates
• Architecture
• Monitoring
• Changelog
• Public roadmap
• System status

Contributing

Lago is built in the open. Read the contributing guide and development environment setup to get started.

Look for issues labeled beginner or help-wanted, or join the Lago Slack community.

License

The Lago platform is distributed under the AGPLv3 license. Read why Lago chose AGPLv3. Lago's Agent SDKs and MCP server are distributed from separate repositories under the MIT license.

Analytics and tracking

Self-hosted Lago instances collect basic product analytics by default. Lago does not collect customer PII or financial data through this tracking. Review what is collected or opt out.