Realm

gnomcp

gno.land/r/g15x77qw5mfjldz59suhg5q5ayau98j8emcjes6j/gnomcp

← All realms

Render

gnomcp

gnomcp is an MCP server and agent skill that connects gno.land to your coding agent: Claude Code, Claude Desktop, Cursor, Gemini CLI, OpenCode, or anything else that speaks MCP. Once it's installed you talk to your agent in plain language, and it reads realms, checks accounts, deploys code, and sends transactions on a real chain.

[!WARNING] Work in progress, unaudited. Writes only work on dev and testnet chains. Mainnet and betanet stay read-only.

Why gno.land works well for agents

Everything on gno.land is readable Go-like source stored on chain, with transparent state. An agent can read a realm the same way it reads any codebase. What was missing was a safe way to let it act. gnomcp adds that part: live chain queries, a testnet key of its own to sign with, and a bundled gno skill that teaches it where Gno differs from Go, so it stops guessing from Go or Solidity habits.

Explore

Ask what realms exist, what they render, and what an account holds. Live reads work on any chain, including this one.

Build

Describe what you want. The agent writes the realm, tests it locally, does a security pass, deploys to testnet, and calls it to prove it works.

Audit

Ask for a security review of any deployed realm. Read-only, works on mainnet too, findings backed by quoted source.

Try it

Once gnomcp is installed, tell your agent:

"Deploy a check-in board at gno.land/r/test/checkin: anyone can check in, it records their address, and reading it back lists everyone so far. Then check in yourself."

New to gno.land? Ask it to teach you instead, it will walk you through a hands-on tour.

Install

One command installs the server binary and the agent skills:

curl -fsSL https://raw.githubusercontent.com/gnoverse/gno-mcp/main/scripts/install.sh | sh

This pipes a script from the internet into your shell, so read it first: scripts/install.sh. Then restart your editor or agent.

Safety

  • Your keys stay in gnokey. The agent signs with its own throwaway testnet key. Acting as you always goes through an authorization you sign yourself, on your own machine.
  • No signing on real-funds chains. The mainnet write path does not exist in the code.
  • Chain output is data, not instructions. Everything read from the chain is wrapped as untrusted content, so a malicious realm can't steer the agent.
  • Every write is logged, and the log names which account signed.

Links

Source (qfile)

gnomcp.gno
gnomod.toml