> ## Documentation Index > Fetch the complete documentation index at: https://docs.weaveos.com/llms.txt > Use this file to discover all available pages before exploring further. # Router billing and credits > How prepaid credits, subscriptions, spend limits, and bring-your-own-key billing work Weave Router turns are funded in the following order, decided per request: | Funding source | When it applies | Effect on shared credits | | - | - | - | | Your compatible linked subscription | A Boost turn can run on linked Claude or ChatGPT (Codex) capacity | None | | Your included seat allowance | An assigned Max or Boost seat has room in both its six-hour and billing windows | None | | Organization Router credits | Eligible paid fallback runs through Weave's provider accounts | Debited at the upstream cost of the turn | | Your own provider account (BYOK) | The turn runs on a provider key you added | Debited for Weave's 5% platform fee only | Each Max or Boost seat has its own included allowance. Allowance is not a purchased balance and cannot be transferred between members. Purchased credits belong to the organization and are shared by all eligible requests. Max remains limited to open-source models when a request uses shared paid fallback. ## Prepaid credits Credits are prepaid and organization-wide. Admins manage them under **Billing & spend → Balance & auto-recharge** in [router settings](https://app.workweave.ai/organization/settings/weave-router). Members see their own allowance and whether shared funding is available, not the shared balance or organization ledger. * Each paid turn debits the upstream cost of that turn — Weave's margin is taken on top-up, not per call. * Self-serve top-ups run from **\$5** to **\$10,000**. Larger commitments and invoicing go through your Weave contact. * A top-up fee of 5.5% (minimum \$0.80) is added to the charge, so a \$100 top-up charges \$105.50 and grants \$100 of balance. * The **Recent activity** ledger shows every top-up and adjustment. ### Auto-recharge Auto-recharge is off until an admin explicitly enables it. Save a card and set **Recharge when below** plus the amount to add, and Weave tops the balance up automatically. An optional **Monthly max** caps how much new credit auto-recharge can purchase in a calendar month; it does not limit use of credits already in the wallet and does not reset anyone's included allowance. Existing organization auto-recharge settings remain in effect until an admin changes them. ### Paid fallback Paid fallback allows eligible requests to use shared organization credits after compatible linked capacity and included allowance are unavailable. It defaults on, but an admin can turn it off under **Billing & spend**. Turning paid fallback on never enables auto-recharge or authorizes a card charge; the wallet must already be funded or separately configured for auto-recharge. Losing a seat or included subscription entitlement stops included coverage but does not by itself revoke Router access. An otherwise-authorized active organization member may continue with shared credits as pay as you go. Organization membership, valid credentials, paid-fallback policy, model restrictions, and organization, engineer, and key spend limits still apply. ### When credits run out What happens depends on whether the turn can run on a subscription: * **A subscription is available** — the turn is served on your own Claude or ChatGPT plan and the response is prefixed with a warning that credits are depleted and paid fallback is disabled. * **No subscription is available** — the request fails with `402 Your organization's prepaid credits are depleted. Contact your org admin to add credits.` Add credits at [Organization settings → Weave Router](https://app.workweave.ai/organization/settings/weave-router), or turn on auto-recharge so it doesn't happen again. ## Spend limits Under **Billing & spend → Spend limits**, admins can cap: * **Per-engineer monthly limit** — the default ceiling for each engineer's router spend * **Org monthly limit** — a ceiling across the whole organization Individual keys can also carry their own spend cap, set when the key is issued. Limits bound **paid** spend, and they are hard stops rather than alerts. Once one is reached: * Turns that would run on a paid model are refused with a 402 and a message naming the limit that was hit. * Turns your own Claude or ChatGPT subscription can cover keep flowing, served on that subscription with paid fallback disabled. After the org limit or a per-key cap this holds in every subscription routing mode except **Use prepaid credits only**. After a per-engineer limit it holds only when subscription routing is set to **Strict pass-through until low**. If an engineer is blocked but still has plan headroom, raise their limit, or have them run `npx @weave-os/router off --claude` (then quit and reopen the client) to use their subscription directly in the meantime. ## Bring your own provider keys Add keys under **Billing & spend → Your provider keys** to run inference on your own Anthropic, OpenAI, Google, or other provider accounts. Provider spend then lands on your provider bill, and Weave debits a **5% platform fee** on that spend from your credit balance — so an org on bring-your-own-keys (BYOK) still needs a funded balance, and the **Use my provider keys** toggle won't turn on without one. The provider list offers only the providers the router can currently route to for your org, and one key per provider is active at a time — adding a second key for the same provider replaces the first. Keys are encrypted at rest and only decrypted to sign upstream calls. Only org admins can view or change them, and the raw key is never shown again after you save it — the list shows a prefix and suffix only. ### Custom endpoints If you run your own deployment of a provider — a gateway, proxy, or private-cloud endpoint that speaks the provider's API — point a key at it with **Advanced → Custom endpoint** when you add the key. Requests authenticated with that key go to your endpoint instead of the vendor's. Leave it blank for the vendor's standard endpoint; the **Endpoint** column shows `Custom` or `Default` per key. Give the same base URL your provider client would use, for example `https://gateway.example.com/v1`. The endpoint must be: * **HTTPS** — plain `http` is rejected * **Publicly resolvable** — loopback, private, link-local, and carrier-grade NAT addresses are rejected, so an endpoint reachable only from inside your network won't work * **Free of embedded credentials** — `https://user:pass@host` is rejected; the provider key you add is the credential A custom endpoint changes where the key's requests go, not how they are billed: the turn still bills your provider account, and Weave still debits the 5% platform fee. An enterprise gateway such as Snowflake Cortex — one that serves several vendors' models under its own model names, or wants a key pair instead of a token — is set up the same way, with a few extras covered in [custom endpoints and gateways](/router/custom-endpoints). ## Reporting lag Router spend and savings normally appear within about an hour. If numbers look stale beyond that, see [troubleshooting](/router/troubleshooting). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.