Taxes, if applicable, appear at checkout. Text and tool-call output; image input depends on the model.
Purchase, receive your key, connect your agent
One private API key for eight models. Seven routes support image input; GLM-5.3 is text-only. Shared 5-hour and 7-day points allowances apply. Payment is processed by Waffo; your key appears in Billing & API keys after verified payment.
Choose a payment option and click Subscribe or Buy 1 month to open secure checkout. Sign in if prompted; checkout continues automatically. After verified payment, a pre-generated key is assigned and activated on the server. Its service period starts at activation. Find your key, access period and usage in Billing & keys.
All models share both points allowances. Each request counts against the 5-hour and 7-day limits; these allowances cannot be added together. Model rates determine how many tokens the balance provides. Fixed windows start when your key is first activated on the server. Renewing does not reset the windows early. Unused points do not roll over. Each key allows up to 4 concurrent requests and 200 requests per minute.
Points count input and output tokens at each model's rate. Both windows apply. The server rejects requests that exceed your allowance. See your account for usage and reset times.
Models and setup
Use your delivered API key with the setup guides below. Each includes the service URL, protocol, configuration files, model selection and reasoning controls.
Check that the models, allowance and setup match your workflow. SaveMyToken provides this API access service.
Who it suits
Developers who want to switch between 8 models in a compatible coding agent using one shared allowance.
What you receive
An assigned API key, service URL, model IDs and usage view in Billing & API keys. API access starts at server activation.
How usage counts
500M points per 5 hours; 2,800M points per 7 days. Points = (input + output tokens) × the model’s rate. Unused allowance does not roll over; rate and concurrency limits also apply.
Check your tool
Follow the setup guide for your agent and version. API access is separate from an IDE subscription. Check image support, tool calls and any native features you need.
API data handling
Content your tool includes in a request passes through SaveMyToken and the upstream inference service. Usage records are separate from request content. Review the data-handling policy for your project’s requirements.
Choose how to pay
Subscription renews monthly until canceled. Pay once provides one month without automatic renewal. Check the selected price and any tax before paying.
Cancellation and help
Cancel future renewal in Billing & API keys. Cancellation is separate from a refund request. For delivery or billing problems, contact support with your order number.
Multi-Model Coding Plan setup: install, add your delivered connection, declare models, choose effort and verify a file-reading task. Each guide includes configuration downloads and desktop notes.
Choose a tier and payment option, then click Subscribe or Buy 1 month to go directly to secure Waffo checkout. If you need to sign in, checkout continues automatically afterward. After verified payment, Billing & keys shows your orders, private API key, API URL, access period and remaining points. Manage or cancel subscription renewals there, including when new purchases are paused.
What is the difference between Pay once and Subscription?
All plans provide 1 month of access per paid period. Pay once does not renew automatically. Subscription renews every month. A renewal period is separate from the 5-hour and 7-day allowance windows.
When does my access start?
After verified payment, your assigned key is activated on the server and its API access period begins. Pre-generating a key does not start access; waiting to make your first API request does not delay activation. Subscription charges follow the payment provider's billing schedule. Check Billing & API keys for your access dates and subscription details.
Do models each get their own allowance?
No. All models share the 5-hour and 7-day point balances. Input plus output tokens are multiplied by the model rate and charged to both windows. The two balances are not added together; either limit can pause usage.
When do limits reset? Does renewing reset them?
Fixed 5-hour and 7-day windows start when your key is first activated on the server. Billing & API keys shows each reset time. Renewals preserve your key and original window anchor; paying does not reset either allowance early. Unused points do not roll over.
Can I send images or generate media?
Seven Multi-Model routes accept text and image input; GLM-5.3 accepts text only. Outputs are text or tool calls. Image input uses your shared allowance and requires a compatible client configuration. Audio and video input and image, audio or video generation are not included. Check the model table for each route's input support.
How do I cancel or request a refund?
Open Billing & API keys → Manage subscriptions to cancel future renewal. A canceling subscription retains access until its paid period ends. For a refund request or a delivery problem, contact contact@savemytoken.com with your order number. A full refund revokes that purchase's API access.
Where do my prompts and code go?
Content included by your coding tool is processed by the SaveMyToken gateway and the upstream inference service for the requested route. The billing usage ledger records identifiers and usage rather than the prompt or response text. Your tool, operational logs and the upstream service have separate data-handling rules. Review our API data-handling details and contact us if your project has specific retention, training-use or region requirements.
Can I use one configuration for every agent?
Each agent has its own schema, credentials and model picker. Start with its guide below. Codex uses Responses, Claude Code uses Messages, and the other examples use Chat Completions. Desktop applications may need separate credentials or a compatibility endpoint.
Why do some configuration values contain null?
In a Pi or OpenClaw thinkingLevelMap, null disables an unsupported choice. It is not an effort sent to the model. Defaults must select a supported value. An Off control sends none where the gateway supports it; omitting a parameter can leave reasoning on.
Does a gateway model include every native agent feature?
Model access does not by itself include native web search, image support or every provider-specific tool. Confirm those features on the delivered route. The setup guides include a separate file-reading check and instructions for diagnosing tool failures.
