Skip to content
ChainFoundry

MCP server · In development

Your AI agent will soon read every blockchain.

ChainFoundry's MCP server gives AI agents decoded, structured data from 7 architectures and 500+ networks — self-hosted, no API keys. Preview coming soon.

Coming soon — the MCP server is actively in development.
The underlying primitives (chaincodec, chainrpc, chainindex) already ship across 7 architectures. The MCP wrapper is next. Join the waitlist to get early access.
Get early access

What is MCP?

Model Context Protocol (MCP) is the open standard — created by Anthropic — that lets AI agents connect to external tools and data. ChainFoundry's MCP server makes blockchain data a native capability for any MCP-compatible AI system.

Under the hood it's just our regular SDK — chaincodec, chainrpc, chainindex, and chaincorrelate — exposed as a set of tool calls the agent can reason about. One config, any chain, every framework.

No API keys
Self-hosted
Any MCP client
Rust + Python
7 architectures
MIT licensed
Preview
bash
pip install chainfoundry-mcp   # coming soon

Claude Desktop config (claude_desktop_config.json):

json
{
  "mcpServers": {
    "chainfoundry": {
      "command": "chainfoundry-mcp",
      "args": ["--chains", "ethereum,solana,cosmos,sui"]
    }
  }
}

Then in Claude: “Decode the last transfer event on Ethereum transaction 0x…” or “Did this wallet bridge anything to Solana in the last hour?”

One agent, every chain

How a multichain conversation actually flows.

A single agent, reasoning across heterogeneous architectures, with no per-chain API integration work.

user
Has 0xA1B2…9D moved any funds out of Ethereum in the last 24h, and did they arrive on Solana?
agent → tool
calling get_transaction(chain="ethereum", address="0xA1B2…9D", since="24h")
agent → tool
calling correlate_bridge(source="ethereum", dest="solana", from="0xA1B2…9D")
agent
Yes — one outbound transfer of 480 ETH to Wormhole on Ethereum, matched to a wrapped ETH mint on Solana 9m later. Time delta and amounts check out. Want the full correlation graph?
Under the hood: two MCP tool calls, dispatched in parallel, each decoded against the chain's native encoding (ABI on EVM, Anchor IDL on Solana, Wormhole VAA matching across both).

Tools your agent gets

8 primitives. Every architecture.

Every tool is backed by the same battle-tested primitives used by production ChainFoundry SDKs — so the agent sees one schema, not seven.

MCP toolWhat it does
decode_eventDecode a transaction's events on any chain
decode_callDecode a function / instruction / message call
get_blockGet block header + tx list from any chain
get_balanceNative + fungible token balances at any height
get_transactionFull tx details with decoded events + errors
correlate_bridgeTrace a bridge deposit to its withdrawal on another chain
list_supported_chainsShow all supported chains and architectures
list_schemasBrowse bundled protocol schemas (CSDL)

Multichain by construction

The same agent call, translated per chain.

The MCP server normalizes every chain's binary encoding into a single canonical shape — so your agent can reason about a Sui object change the same way it reasons about an ERC-20 Transfer.

Ethereum
decode_event → Uniswap v3 Swap
Solana
decode_event → Raydium swap (Anchor IDL)
Cosmos
decode_call → Osmosis IBC transfer
Polkadot
get_block → SCALE-encoded extrinsics
Bitcoin
get_transaction → UTXO graph + Ordinals inscription
Sui
decode_event → Move object mutation
Aptos
get_balance → Move fungible asset store

How it works

Five hops. Every chain.

01
User
Prompt from Claude Desktop, Cursor, or any MCP client
02
AI agent
Decides which tool to call and with what arguments
03
MCP protocol
Standard JSON-RPC transport (stdio / WS)
04
ChainFoundry MCP
Dispatches to chaincodec, chainrpc, or chainindex
05
Chain nodes
EVM · Solana · Cosmos · Substrate · Bitcoin · Aptos · Sui