SandBase CLI: Connect Your AI Agent to 2,000+ Models and APIs

Introduction

The @sandbaseai/cli is an open-source command-line interface and local Model Context Protocol (MCP) server designed to supercharge your AI agents. It provides a seamless bridge, connecting 25 popular AI clients, including Claude Code, Cursor, Codex, ChatGPT, Hermes, and OpenClaw, to an expansive ecosystem of over 2,000 AI models and APIs. This powerful tool simplifies complex integrations with features like OAuth and automatic rollback, ensuring a secure and reliable connection.

Traditionally, AI coding assistants are limited to their native capabilities, unable to browse the web, interact with social media, generate images, or fetch real-time data without manual API integration. SandBase CLI revolutionizes this by offering a unified command to connect your agent to a vast array of AI models and APIs via the Model Context Protocol (MCP).

Why Use It & Key Benefits

SandBase CLI addresses common pain points in AI agent development, offering significant advantages:

Simplified Integration: Connect to 2,000+ AI models and APIs with a single command, eliminating the need to manage individual API keys or complex configurations.
Broad Client Compatibility: Supports 25 AI clients, ensuring your preferred coding assistant or agent can leverage its capabilities.
Verifiable Agent Workflow: Guides your agent through a structured process for discovering, inspecting, and running models. This includes tools like sandbase_discover to find suitable models, sandbase_inspect to check schemas and pricing, and sandbase_run to execute tasks.
Enhanced Security: Implements robust security measures such as OAuth Device Flow with PKCE, restricted file permissions for credentials, automatic rollback on errors, and the ability to revoke access anytime via the SandBase Dashboard.
Open-Source Ecosystem: Part of a larger open-source stack, including SandBase Harness for self-hosted agent runtime and SandBase Agent Skills with 88 installable skills.

Installation

Getting started with SandBase CLI is straightforward.

You can use npx for a quick connection:

npx -y https://github.com/sandbaseai/cli/releases/download/v0.1.17/sandbaseai-cli-0.1.17.tgz connect

Alternatively, install the official Homebrew formula on macOS or Linux:

brew install sandbaseai/tap/sandbaseai-cli
sandbase connect

Once connected, your agent gains immediate access to the extensive SandBase ecosystem.

Examples

SandBase CLI provides a clear, verifiable workflow for your AI agent. After connecting, instruct your agent to follow these steps:

Use sandbase_discover to search for suitable models or APIs.
Use sandbase_inspect to review input schema, current prices, and requirements.
Confirm the endpoint, parameters, and potential costs before executing sandbase_run.
Track asynchronous calls with sandbase_run_get using the returned run_id.
Check the status and accrued costs of recently executed calls with sandbase_runs.
Verify your current account balance with sandbase_account.

For instance, you can begin with a non-billable search query:

Find image models for a square product illustration. Compare inputs and current prices of the two best candidates. Do not execute any model yet.

Here are some core commands:

sandbase connect [--client <name>]    # Authorize + configure
sandbase doctor [--client <name>]     # Health check
sandbase unregister [--client <name>] # Remove configuration
sandbase catalog --json               # List supported clients

And the MCP tools available after connection:

Tool	Purpose
`sandbase_discover`	Browse 2,000+ models and APIs
`sandbase_inspect`	Retrieve schema, pricing, and template
`sandbase_run`	Execute model or API
`sandbase_run_get`	Query status of asynchronous tasks
`sandbase_runs`	View recent API calls and costs
`sandbase_account`	Check account balance (free)