Quickstart

Let your AI assistant help you build

Paste this into your AI assistant's chat, then describe what you want to build or share the SandBase API page you are using. This is an instruction for your assistant, not a terminal command.

set up https://www.sandbase.ai/SKILL.md

The setup Skill guides your agent through your current situation: no account yet, an existing API key, an already-connected tool, an existing project, or a new feature. It checks the endpoint's inputs and price, helps implement the integration, and distinguishes setup from a verified call and working feature. Tell it your goal and any spending limit; do not paste your key into chat.

Already have a key or a project? Say so; you do not need to start over. For example: “I have a SandBase key. Add a Similarweb traffic lookup to this project; inspect the API and show the live-test cost first.” An assistant that cannot execute commands or load tools needs the client-specific setup path.

Give your agent the Markdown Quickstart, or follow the account, website, direct API, and hosted Agent paths below. Installing the setup Skill does not itself create a hosted Agent.

Onboarding3 steps

Create an account, set up an organization, then make one real request.

Create accountCreate organizationCall an API

Already have an account? Skip to create an organization or connect SandBase .

1. 01Create your account

   Sign in with GitHub or Google. First login creates your account and a personal workspace.

   Open login
2. 02Create an organization

   Keep the personal workspace, or create a team so keys, members, and billing are shared.

   Open Team
3. 03Connect SandBase

   Call an API, try it on the website, connect an AI tool, or build an Agent.

   See paths

Choose a path

ACall an APISend a request from your appBUse on the websiteTry it in the Store firstCUse in an AI toolConnect the client you already useDBuild an AgentCompose a reusable workflow

1. Create your account

SandBase uses GitHub or Google. There is no separate email-and-password signup.

1. Open Create account / Sign in.
2. Complete the security check, then choose Sign in with GitHub or Sign in with Google.
3. Approve access in the provider window.
4. Confirm you land in the Console.

First login creates your account and a personal workspace. You do not need to create an organization before you can try SandBase.

If a welcome credit appears, you can start without adding a payment method. The Balance card on the Console Dashboard and Billing show the current balance.

2. Create an organization

API keys, usage, and billing belong to an organization. Stay in the personal workspace if you are evaluating alone. Create a team when several people will share keys and spend.

Workspace	When to use it
Personal	Created automatically on first login. One per user.
Team	Shared members, keys, and balance. Create this for a company or project.

2.1 Create a team

1. Open Create team, or choose Create Team in the account menu at the top right of the Console.
2. Enter a Team name, then select Create team.
3. Confirm the account menu now shows that team as the active workspace.
4. On Developer → Team, select Add member, enter the Account email of an existing SandBase account, and choose the Workspace role (Member or Admin).

An invited person must already have a SandBase account. If they do not, send them this page so they can sign in first, then add them.

If you belong to more than one organization, switch in the account menu. Keys and billing follow the active organization. See Organizations for roles.

3. Connect SandBase

Pick the path that matches the work you need to do now. You can add the others later. Confirm the active organization first so keys and connected tools land in the right workspace.

3.1 Call an API

Choose this if you are adding one capability to your app.

1. Create an API key in Developer → API Keys with New key. Copy it immediately; it is shown only once.
2. Filter the catalog and open a detail page:
   • Language models — filter by provider, capabilities, context, and price
   • Image and video models — filter multimodal models such as text-to-image, image-to-image, text-to-video, and image-to-video
   • APIs — filter API resources by platform and data source
3. Copy the request and call it with your SandBase API key.

Example Model call:

export SANDBASE_API_KEY="sk-YOUR_KEY"

curl https://api.sandbase.ai/v1/chat/completions \
  -H "Authorization: Bearer $SANDBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Summarize this customer feedback."}]
  }'

Learn more: First API Call.

3.2 Use SandBase on the website

Choose this if you want to try a Model, API, or Agent in the browser first — no client install and no code.

1. Sign in, then open the catalog:
   • Language models
   • Image and video models
   • APIs
   • Agents
2. Filter and open a detail page.
3. Run one request on that page and confirm a result comes back.

3.3 Use SandBase in your AI tool

Choose this if you use Claude Code, Codex, Cursor, Kiro, or another supported client and want SandBase Models and APIs inside that app.

1. Open Build → AI tools in the Console.

2. Paste this line into your AI assistant's chat, then tell it what you want to build:

   set up https://www.sandbase.ai/SKILL.md

3. Approve the browser sign-in the agent opens. There is no API key to paste.

4. Start a new session or restart the app so the SandBase tools load.

5. Ask the agent to discover and inspect the relevant tool, then make one bounded request within your authorization and budget. For development, have it integrate the response into your project and provide the run command and actual test results. Discovery alone does not verify a business API call.

Prefer a terminal? Run npx -y @sandbaseai/cli@latest connect. The connected tool can search the whole catalog and runs are billed to the active workspace.

For example, you can ask:

Research this company and give me a sourced customer brief.

Learn more: Connect AI tools.

3.4 Build an Agent

Choose this if the work has multiple steps or should be reused.

1. Open Build → Agents and select Create.
2. Fill in Overview and Instructions.
3. Pick a Model.
4. Add Skills and MCP tools under Capabilities, then set Runtime.
5. Select Create & Chat to test it in a Session.
6. Call the tested version from your app on the Agent's API tab.

Learn more: Build Agent. To do the same from code with the official OpenAI SDK, see OpenAI compatibility.

4. Confirm it works

You are ready when all of the following are true:

• The Console account menu shows the organization you intend to use.
• You completed one path: an API call, a run on the website, a connected AI tool, or an Agent.
• One request succeeds.

If the request fails, check the active organization, the key, and Errors. Top up in Billing when the workspace has no remaining balance.

5. Billing and usage

After you sign in, review spend for the active organization:

BillingBilling and credits

Check balance, top up, and review billing records.

AnalysisUsage analysis

See recent call volume, cost, and request health for the active organization.

6. What to do next

ModelsLanguage models

Filter the LLM catalog by provider, capability, context, and price.

MultimodalImage and video

Filter text-to-image, image-to-video, and other visual models.

APIsAPI resources

Filter search, social, and data APIs by platform.

• API keys — create and rotate organization keys
• Models and APIs catalog — try Models, APIs, and Agents on the website
• Connect AI tools — give your AI tool the SandBase catalog
• Build Agent — create reusable Agents
• API Reference — integrate with code