Sign in Subscribe

Portal MCP Server: Connect AI Agents to Onchain Data

mojo

• 2 min read

A lot of onchain work still starts the same way: you’ve got a question, and then you end up doing the boring setup work just to get to the first result. We just made that process faster.

SQD Portal API is now available as a remote MCP server: portal.sqd.dev/mcp

0:00
/0:22

If you use Claude, Cursor, or VS Code, you can connect Portal as a tool source and query onchain data from inside your AI workflow. Instead of “write integration first”, you can simply ask a question and the AI can call structured Portal tools directly.

Faster build loop

You can iterate on questions without building everything from scratch:

  • Use timeframes like “24h” / “7d” instead of manual block maths
  • Start with higher-level tools (wallet summary, recent txs, transfers)
  • Drop to raw logs / traces only when you need precision

The tool surface is big enough to build real features

Portal MCP ships 40+ tools across 225+ datasets, grouped into:

  • Discovery (portal_get_sqd_info, portal_list_datasets, portal_search_datasets)
  • Core queries (EVM logs/txs/traces/blocks/state diffs + Solana instructions/balances/logs/rewards)
  • Transfers (ERC20 + NFTs + aggregation)
  • Analytics (wallet summary, time series, gas analytics, contract activity, top contracts/addresses, count events)
  • Multi-chain (batch query, compare chains, block-at-timestamp, decode logs, paginated queries)

Multi-chain becomes normal

You can run the same question across chains without building separate pipelines using portal_batch_query and portal_compare_chains.

A quick demo prompt (after connecting the MCP server)

“Give me a wallet summary for 0x… on Base for the last 7 days.

Show tx count, ERC20 transfers in/out, NFT transfers, top interacted contracts, and a daily activity trend.

If it’s large, summarise and show the top 10 per section.”

For years, SQD has been building for a world where AI agents are normal. We’re already seeing more automated behaviour onchain, and the next wave will likely involve a much greater demand for data than most people expect. That raises the bar for indexing infrastructure: it needs to scale quietly and consistently.

That’s the direction Portal has been built for: decentralized by design, and now agent-ready via MCP.

Try it

Docs page: beta.docs.sqd.dev/en/ai/mcp-server-portal

Server URL: portal.sqd.dev/mcp

If you build anything that needs onchain data - analytics, research tools, monitoring, growth dashboards, or agent workflows - this should dramatically improve your workflow.