Ask Cat › AI Tool Summary › SandBase CLI
SandBase's Pricing Page Has No Numbers on It: Where the Free Credits Start and the Five Steps to Your First Call
Article last updated:2026-09-07
SandBase puts models, third-party APIs and agents on one platform. Its pricing page has an unusual property: there is not a single monetary figure on it.
That is not a broken page — it follows from how the product bills. Here is how the money works, where free starts, and the official five-step path to a first call. Verified 2026-09-07.
1. What the pricing page actually says
The stated principle is “Pay for delivered work, not idle infrastructure.” Three tiers:
| Tier | As described |
|---|---|
| START | Free credits — explore Models, APIs and Agent workflows with starter credits, no subscription |
| BUILD | Pay as you go — pay for the underlying model, API and runtime work your calls use, usage based |
| SCALE | Team / Enterprise — centralised workspaces, controls and support, contact us |
No monthly fee and no tiered rate card, because cost is decomposed into three components:
- Model work: token, image, video or audio generation, priced by the selected provider model.
- Real-world APIs: usage follows the published pricing of the API capability you call.
- Agent runs: component work and runtime work shown together in each traceable execution.
So the answer to “what does it cost” is not on the pricing page — it is on the page for the model or API you pick. Honest, but it means you have to browse the Store, where capabilities, pricing, input schemas and working examples are listed.
2. Two ways to build
The docs open with the choice, advising you to start with the smallest surface that fits and combine later:
| Models and APIs | Managed Agents | |
|---|---|---|
| What it is | Direct access to language, image, video, audio and third-party APIs | Reusable, versioned Agent configurations executed by SandBase |
| Best for | Custom application loops, fine-grained request control | Multi-step, asynchronous, scheduled or repeatable work |
| You manage | Prompts, orchestration, retries, application state | Agent definition, enabled capabilities, inputs, delivery behaviour |
One API key calls Models, APIs, Agents, Services and scheduled workloads. The docs also state it plainly: keep keys on the server and never expose them in browser code.
3. Four terms, one hierarchy
SandBase uses four words that are easy to conflate:
- Agent: a reusable, versioned configuration containing a model, instructions, APIs, Skills and credentials. Define once, test, then choose how it runs.
- Service: exposes a tested Agent to applications. Its compatibility API path is
/v1/endpoints; invoking it creates or continues a Session. - Session: one durable interaction, identified by
session_id. - Schedule: executes an Agent later or repeatedly. Its API resource is a Deployment; every trigger creates a distinct
drun_*DeploymentRun and, on successful Session creation, links it to a newsess_*Session.
Plainly: the Agent is the recipe, the Service is the counter you open for it, the Session is one conversation, and the Schedule is what knocks on the counter for you.
4. The official first steps
The documented order:
- Create an organization and API key.
- Choose a Model, API or Agent from the Store (which shows capabilities, pricing, input schemas and examples).
- Make a direct request, or define and test an Agent.
- Publish the Agent as a Service when an application needs to invoke it.
- Add a Schedule when it needs to run repeatedly.
The order matters: confirm behaviour and cost with a direct request, then wrap it as an Agent, and automate last. Doing it backwards — schedule first, test later — is the standard way automation burns money.
5. Three things to do while on free credits
- Check the Store price before you call. Billing follows the model you select, so swapping one model in the same code can change cost several-fold. The docs state the Store shows pricing.
- Run direct requests before creating a Schedule. Every trigger is a real execution; the
drun_*records will tell you how many happened, but the spend is already committed. - Keep keys server-side. The docs call this out explicitly, which usually means it is a common mistake.
6. What this article could not verify
- Neither the pricing page nor the docs home publishes a figure or credit count for the free tier, so we do not estimate one. The exact allowance appears after you create a workspace, on the billing page.
- The docs’ Pricing, Billing & credits and Rate Limits sub-pages load interactively and could not be read directly; this article is based only on the directly readable pricing page and docs home.
For comparison with another usage-billed platform’s free allowance, see OpenRouter’s daily free model limits.
Sources: SandBase’s pricing page and documentation home, read directly on 2026-09-07. Quoted descriptions are translations of the official copy; the free credit amount is flagged as unverified because SandBase does not publish it. Plan summary also on our SandBase CLI tool page.
Let's take a look at these
- SandBase CLI Comprehensive Introduction: Pricing, Features, and Actual Limitations
- SandBase CLI Is the free quota enough?
- SandBase CLI Alternatives
- Comprehensive Free Quota List for All Tools

