use-sapiom

name: use-sapiom description: Access 97 cloud-service tools via Sapiom MCP server — scrape websites, run code in sandboxes, store data (Redis/Vector/Postgres), search the web, call LLMs, generate images and audio, send messages, and manage agent governance. Use when user says "use sapiom", "scrape this page", "search the web", "run this code in the cloud", "store data", "generate an image", "create a database", "set up a message queue", or needs any paid cloud service without vendor account setup. Do NOT use when sapiom-deploy or sapiom-agent-builder skills are loaded (those are more specific).

Sapiom

97 cloud-service tools via one MCP server. Scrape, compute, store, search, generate — pay per use, no vendor accounts.

Quick Setup — MCP Server (Recommended)

# Claude Code CLI
claude mcp add sapiom --transport http https://api.sapiom.ai/v1/mcp

{
  "mcpServers": {
    "sapiom": {
      "type": "http",
      "url": "https://api.sapiom.ai/v1/mcp"
    }
  }
}

Browser auth is used automatically. To use an API key instead, add --header "Authorization:Bearer YOUR_KEY" or set the headers field. Get a key at https://app.sapiom.ai/settings.

When to Use

• Scrape & extract web content, crawl sites, take screenshots
• Run code in cloud sandboxes (Node, Python, Go, Rust) with file I/O and process management
• Store data in Redis, Vector DB, Search Index, or ephemeral Postgres
• Search the web with Linkup (fast) or You.com (thorough)
• Generate media — images (Fal.ai), speech (ElevenLabs), sound effects
• Send messages via queues, schedules, and pub/sub

Building an agent? Use sapiom-agent-builder. Just deploying code? Use sapiom-deploy.

Available Tools

Getting Started (2)

Scraping & Browser (9) — details

Compute (31) — details

Data (23) — details

Messaging (16) — details

Search (2) — details

AI Models (2) — details

Audio (3) — details

Images (1) — details

Verify (2) — details

Governance (6) — details

Recipes

1. Research a Topic

sapiom_search (quick answer) → sapiom_deep_search (more results) → sapiom_fetch (read full articles) → sapiom_chat (summarize findings)

2. Monitor a Page for Changes

sapiom_scrape (get content) → sapiom_redis_command (store hash, compare with previous) → sapiom_message_publish (alert on change) → sapiom_schedule_create (run every hour)

3. Build a Knowledge Base

sapiom_crawl (crawl docs site) → sapiom_crawl_status (get pages) → sapiom_vector_upsert (store embeddings) → sapiom_vector_query (semantic search)

4. Run & Deploy Code

sapiom_run (test code snippet) → sapiom_sandbox_create (persistent environment) → sapiom_sandbox_exec (iterate) → sapiom_job_deploy (schedule for production)

5. Generate Media

sapiom_generate_image (create visuals) + sapiom_text_to_speech (narration) + sapiom_sound_effects (background audio)

Tips

• tool_discover finds the right tool by keyword — it's free, use it when unsure
• Pass agentName on gateway calls for per-agent spend tracking
• sapiom_search (Linkup) = fast AI answer; sapiom_deep_search (You.com) = more raw results
• Provision resources before using them: sapiom_redis_create, sapiom_vector_create, sapiom_database_create
• Set spending rules via sapiom_create_spending_rule before deploying autonomous agents
• sapiom_database_price is free — always check cost before creating a Postgres DB

SDK Access (Direct HTTP)

For direct HTTP calls without MCP, use the SDK:

npm install @sapiom/fetch
# or: npm install @sapiom/axios axios

const { createFetch } = require("@sapiom/fetch");
const safeFetch = createFetch({ apiKey: process.env.SAPIOM_API_KEY });

const res = await safeFetch("https://linkup.services.sapiom.ai/v1/search", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ q: "quantum computing", depth: "standard", outputType: "sourcedAnswer" }),
});

Python/curl: use standard HTTP with X-402-Payment header — see reference files for gateway URLs.

Troubleshooting

References

Full docs: https://docs.sapiom.ai (append .md to any URL for raw markdown)