Polar Python SDK

Use the generated, versioned Polar SDK without inventing methods or parameters.

Inspect the application first

1. Identify the installed polar-sdk version and preserve the API version already selected by the application.
2. Inspect generated service signatures, input TypedDict definitions, output dataclasses, and endpoint-specific errors before writing calls.
3. Use Polar in synchronous code and PolarAsync in asynchronous code. Do not mix blocking SDK calls into an async request path.
4. Identify whether the access token is scoped to an organization. Pass organization_id when the selected token does not imply one, and confirm the required endpoint scopes.
5. Keep client, webhook, and environment configuration on the server.

Migrate from the old SDK

Read Migration from the `<1.0.0` SDK before changing an application that imports polar_sdk, calls _async operation variants, or paginates with .next(). Do not mix old and new client, model, pagination, error, or webhook conventions.

Configure and reuse the client

Keep production and sandbox access tokens separate. Never expose an organization access token in browser, mobile, or other public client code. Pass environment="sandbox" while testing; omit it or pass environment="production" in production.

import os

from polar.v{{ ir.versions[0].version | replace("-", "_") | replace(".", "_") }} import Polar

with Polar(
    os.environ["POLAR_ACCESS_TOKEN"],
    environment="sandbox",
) as polar:
    customer_state = polar.customers.get_state_external("usr_123")

For a long-running service, create one client during application startup, reuse its connection pool, and close it during application shutdown. Do not create a new client for every incoming request.

Use the async client with an application-lifetime async context in async services:

import os

from polar.v{{ ir.versions[0].version | replace("-", "_") | replace(".", "_") }} import PolarAsync


async def load_customer_state() -> None:
    async with PolarAsync(os.environ["POLAR_ACCESS_TOKEN"]) as polar:
        customer_state = await polar.customers.get_state_external("usr_123")
        print(customer_state)

Follow generated argument and output conventions

• Pass path parameters positionally.
• Pass query and request-body parameters as keyword arguments.
• Treat request payload definitions as generated TypedDict schemas.
• Treat JSON object responses as generated, slotted dataclass instances or unions.
• Expect UUID and date-time OpenAPI fields to remain strings in this SDK version.
• Use iter_* for paginated endpoints. Use for with Polar and async for with PolarAsync; each additional page performs another API request.

Correlate customers with stable external IDs

Use the application's immutable user or organization identifier as the Polar customer external_id. It is unique within a Polar organization. It may be assigned later when initially unset, but it cannot be changed or removed once set.

customer = polar.customers.create(
    external_id="usr_123",
    email="user@example.com",
)
customer_state = polar.customers.get_state_external("usr_123")

Handle concurrent creation safely. A preflight lookup followed by creation is not atomic; if creation reports an existing email or external ID, fetch and reconcile the existing customer rather than creating a second mapping.

Create checkout and customer portal sessions

Create checkout sessions on the server and send only the returned URL to the browser. Prefer external_customer_id so a completed checkout is reconciled with the application's customer.

checkout = polar.checkouts.create(
    products=[os.environ["POLAR_PRODUCT_ID"]],
    external_customer_id="usr_123",
    success_url="https://example.com/billing/success?checkout_id={CHECKOUT_ID}",
    return_url="https://example.com/settings/billing",
)
redirect_to(checkout.url)

Do not treat the success redirect as proof of payment or entitlement. Confirm access through customer state or a verified webhook.

Create short-lived customer portal sessions only for an authenticated customer:

session = polar.customer_sessions.create(
    external_customer_id="usr_123",
    return_url="https://example.com/settings/billing",
)
redirect_to(session.customer_portal_url)

Validate return URLs and never let an untrusted caller choose another customer's external ID.

Reconcile feature access from customer state

Prefer a feature_flag benefit attached to the relevant products over inferring access from subscription statuses. Reconcile both access grants and access revocations.

from polar.v{{ ir.versions[0].version | replace("-", "_") | replace(".", "_") }}.outputs import CustomerState


def has_feature_access(customer_state: CustomerState, benefit_id: str) -> bool:
    return any(
        grant.benefit_id == benefit_id
        for grant in customer_state.granted_benefits
    )


customer_state = polar.customers.get_state_external("usr_123")
set_feature_access(
    customer_external_id="usr_123",
    enabled=has_feature_access(
        customer_state,
        os.environ["POLAR_BENEFIT_ID"],
    ),
)

Make set_feature_access an idempotent application write. A missing benefit must disable access, not merely skip a grant operation.

For webhook-driven synchronization, use benefit-grant events for an incremental local grant ledger or customer.state_changed for complete snapshot reconciliation. Read Webhook processing before implementing or modifying a receiver.

Ingest metered usage events

Read Usage-event ingestion before implementing metered billing. In particular:

• Match the event name and metadata to the configured meter.
• Associate the event with exactly one customer_id or external_customer_id.
• Assign every logical event a stable, unique external_id.
• Reuse the same external_id when retrying so Polar can deduplicate the event.
• Inspect both inserted and duplicates in the ingestion result.
• Enforce usage limits in the application; ingestion does not reject an action because a customer has exhausted a balance.

Handle API failures intentionally

Handle generated endpoint-specific errors before broad SDK errors. Classify common base errors as follows:

• PolarNetworkError: retry only when the operation is safe or carries a stable deduplication identifier.
• PolarRateLimitError: apply bounded backoff and honor retry_after when present.
• PolarServerError: retry selected 5xx failures with bounded exponential backoff and jitter.
• PolarClientError: do not blindly retry authentication, authorization, validation, or not-found failures.

Never create a new usage-event external_id during a retry. For other mutations, use an endpoint-supported idempotency mechanism or application-level reconciliation before retrying.