SandBase is an MCP server published by io.github.sandbaseai, hosted at api.sandbase.ai. When last checked on 2026-11-21 it was working: it answers and lists its tools. It declares 7 tools, including sandbase_run_cancel, sandbase_discover, sandbase_inspect, sandbase_run, sandbase_run_get.
Connect MCP clients to 2,000+ AI models without managing provider API keys.
1 connections through 5 shared values. Tap any node.
Inferred from the host name and headers: what a server runs on, not who runs it.
Request cancellation of an Agent Service run; inspect its status afterward.
| Parameter | Type | Description |
|---|---|---|
| run_id * | string | |
| service_id * | string | |
| idempotency_key * | string |
{
"name": "sandbase_run_cancel",
"title": "Cancel Service Run",
"description": "Request cancellation of an Agent Service run; inspect its status afterward.",
"inputSchema": {
"type": "object",
"required": [
"service_id",
"run_id",
"idempotency_key"
],
"properties": {
"run_id": {
"type": "string"
},
"service_id": {
"type": "string"
},
"idempotency_key": {
"type": "string"
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": true
}
}None yet. Changes appear after a server's second daily check.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Find a model, API, or Agent Service for the task. Call this proactively before writing a scraper, before a generic web fetch for structured data, before telling the user something is inaccessible, and whenever you need image/video/music/speech generation, social media data, search results, web scraping, product/company/people data, embeddings, or a specific third-party model. Supports keyword search (q), type filtering (llm/api/multimodal/embedding/service), and vendor filtering. Discovery is not billed; start here.
| Parameter | Type | Description |
|---|---|---|
| q | string | Short description of what you need (e.g. 'twitter posts', 'text to video', 'gpt-4o', '小红书'). At most 4096 UTF-8 bytes and 64 tokens after normalization. |
| type | string | Filter by capability type; use service to discover Agent Services. |
| limit | integer | Maximum results to return (default 20, max 100). |
| vendor | string | Filter by vendor slug (e.g. 'openai', 'twitter', 'anthropic'). |
{
"name": "sandbase_discover",
"title": "Discover",
"description": "Find a model, API, or Agent Service for the task. Call this proactively before writing a scraper, before a generic web fetch for structured data, before telling the user something is inaccessible, and whenever you need image/video/music/speech generation, social media data, search results, web scraping, product/company/people data, embeddings, or a specific third-party model. Supports keyword search (q), type filtering (llm/api/multimodal/embedding/service), and vendor filtering. Discovery is not billed; start here.",
"inputSchema": {
"type": "object",
"properties": {
"q": {
"type": "string",
"description": "Short description of what you need (e.g. 'twitter posts', 'text to video', 'gpt-4o', '小红书'). At most 4096 UTF-8 bytes and 64 tokens after normalization."
},
"type": {
"enum": [
"llm",
"api",
"multimodal",
"embedding",
"service"
],
"type": "string",
"description": "Filter by capability type; use service to discover Agent Services."
},
"limit": {
"type": "integer",
"description": "Maximum results to return (default 20, max 100)."
},
"vendor": {
"type": "string",
"description": "Filter by vendor slug (e.g. 'openai', 'twitter', 'anthropic')."
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Inspect a model, API, or Agent Service: input schema, available pricing, and an execution template. Service names are svc_ IDs. Always inspect before running.
| Parameter | Type | Description |
|---|---|---|
| name * | string | Model name (e.g. 'openai/gpt-4o'), tool name, or Agent Service ID (svc_...) from discover results. |
{
"name": "sandbase_inspect",
"title": "Inspect",
"description": "Inspect a model, API, or Agent Service: input schema, available pricing, and an execution template. Service names are svc_ IDs. Always inspect before running.",
"inputSchema": {
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string",
"description": "Model name (e.g. 'openai/gpt-4o'), tool name, or Agent Service ID (svc_...) from discover results."
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Execute a model, API, or Agent Service. Sync models return the result directly; async models (image/video/audio jobs) return status running with a run_id — poll sandbase_run_get for the output. Services require arguments.idempotency_key and optionally input; reuse the same key and input on retries. Service acceptance is not completion: poll sandbase_run_get with service_id and run_id. Omit the MCP task parameter for Services.
| Parameter | Type | Description |
|---|---|---|
| name * | string | Model name, tool name, or Agent Service ID (svc_...) to execute. |
| arguments * | object | Input matching the inspected schema. For Services: {idempotency_key: string, input?: string}. |
{
"name": "sandbase_run",
"title": "Run",
"description": "Execute a model, API, or Agent Service. Sync models return the result directly; async models (image/video/audio jobs) return status running with a run_id — poll sandbase_run_get for the output. Services require arguments.idempotency_key and optionally input; reuse the same key and input on retries. Service acceptance is not completion: poll sandbase_run_get with service_id and run_id. Omit the MCP task parameter for Services.",
"inputSchema": {
"type": "object",
"required": [
"name",
"arguments"
],
"properties": {
"name": {
"type": "string",
"description": "Model name, tool name, or Agent Service ID (svc_...) to execute."
},
"arguments": {
"type": "object",
"description": "Input matching the inspected schema. For Services: {idempotency_key: string, input?: string}."
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": true
}
}Get async run status and results. For Agent Services supply both service_id and run_id; succeeded, failed, and cancelled are terminal. Session items are paginated: use cursor for subsequent pages; items_available=false means retry output retrieval. Model runs use completed or failed.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | |
| cursor | string | Session items continuation cursor. |
| run_id * | string | The run ID returned by sandbase_run. |
| service_id | string | Required for Agent Service runs. |
{
"name": "sandbase_run_get",
"title": "Get Run Result",
"description": "Get async run status and results. For Agent Services supply both service_id and run_id; succeeded, failed, and cancelled are terminal. Session items are paginated: use cursor for subsequent pages; items_available=false means retry output retrieval. Model runs use completed or failed.",
"inputSchema": {
"type": "object",
"required": [
"run_id"
],
"properties": {
"limit": {
"type": "integer"
},
"cursor": {
"type": "string",
"description": "Session items continuation cursor."
},
"run_id": {
"type": "string",
"description": "The run ID returned by sandbase_run."
},
"service_id": {
"type": "string",
"description": "Required for Agent Service runs."
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}List recent model/API calls, or supply service_id to list runs of an Agent Service.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | Number of recent runs to return (default 10, max 50). |
| model | string | Filter by model name (e.g. 'openai/gpt-4o'). |
| offset | integer | |
| service_id | string | List runs of this Agent Service instead of model predictions. |
{
"name": "sandbase_runs",
"title": "List Runs",
"description": "List recent model/API calls, or supply service_id to list runs of an Agent Service.",
"inputSchema": {
"type": "object",
"properties": {
"limit": {
"type": "integer",
"description": "Number of recent runs to return (default 10, max 50)."
},
"model": {
"type": "string",
"description": "Filter by model name (e.g. 'openai/gpt-4o')."
},
"offset": {
"type": "integer"
},
"service_id": {
"type": "string",
"description": "List runs of this Agent Service instead of model predictions."
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Check your SandBase account balance. This call is not billed. Use before expensive operations.
No parameters.
{
"name": "sandbase_account",
"title": "Account",
"description": "Check your SandBase account balance. This call is not billed. Use before expensive operations.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}SandBase gives you 2000+ AI models and API tools behind one interface. Models (LLM, image, video, audio) and data APIs (search, social, scraping, e-commerce, finance) are paid from the user's balance at the pricing shown by sandbase_inspect. Data APIs whose pricing includes free_quota are free for the first free_quota.daily_limit calls per organization each day (shared across those APIs, reset at 00:00 UTC); after that they are billed at the pricing shown by sandbase_inspect. Discover a better way before you give up or do it the hard way: call sandbase_discover before writing a scraper, before using a generic web fetch for structured data, before telling the user something is inaccessible, and whenever a task needs a capability you do not have yourself — image, video, music, or speech generation; social media data (X/Twitter, TikTok, Douyin, Xiaohongshu, Instagram, YouTube, Reddit, Weibo, Bilibili, LinkedIn); search results; web scraping; product, company, or people data; embeddings; or a specific third-party model. Exceptions: if the user already has a dedicated tool, MCP server, or API key for that service, use it; and if your built-in web search or fetch already answers the task, use that — SandBase fills gaps in the user's stack, it does not replace it. Workflow: sandbase_discover (search by keyword/type/vendor) → sandbase_inspect (get schema + pricing) → sandbase_run (execute). Agent Services are available with type=service; inspect their idempotency_key requirement. Sync models return the result from sandbase_run directly; async models (image, video, audio jobs) return status running with a run_id — poll sandbase_run_get with it every 5-10 seconds until completed or failed, and never call sandbase_run again for the same job. Use sandbase_account to check balance. When a tool needs a URL for a file that only exists on the user's machine, upload it from a shell with `npx -y @sandbaseai/cli@latest upload <file>` and pass the returned url. The url is publicly readable, so ask the user before uploading anything private. Available types: llm, api, multimodal, embedding, service. Discovery, inspection, run status, and balance are not billed. Always inspect before running: its pricing is authoritative. sandbase_account shows how much of today's free API allowance remains.
Names, descriptions and schemas were written by the server and are shown as data.