grimoire-vm

Grimoire

"Verba Volant, Scripta Manent."

Grimoire is a language for agents to express financial intent with readable syntax and deterministic execution. Spells compile to an intermediate representation (IR) and run through protocol adapters, so you can swap venues by changing aliases and configuration instead of rewriting strategy logic.

Examples | Skills

---

Start here

Grimoire has one runtime semantics: preview first, then commit for irreversible actions. The same behavior applies across CLI and library entry points.

For a single onboarding flow for both users and agents, start with:

• docs/tutorials/quickstart-users-and-agents.md

CLI entry point (grimoire)

Use this for reproducible simulation and onchain execution with adapters and state persistence.

npm i -g @grimoirelabs/cli

grimoire setup

grimoire simulate spells/compute-only.spell --chain 1

grimoire cast spells/uniswap-swap-execute.spell \
  --dry-run \
  --key-env PRIVATE_KEY \
  --rpc-url <rpc>

grimoire setup notes:

• setup is always execute-mode onboarding
• guided mode prompts for required values (chain, RPC, wallet source, doctor checks)
• blank RPC input falls back to public chain RPC
• setup runs a smoke preview and venue doctor by default
• when password is entered interactively, setup writes .grimoire/setup.env (can disable with --no-save-password-env)
• CLI auto-loads nearest .grimoire/setup.env on startup and does not override already-set env vars
• set GRIMOIRE_SETUP_ENV_FILE to point to an explicit env file when running outside the project tree
• .grimoire/setup.env contains plaintext secrets; keep it local-only and rotate/delete when not needed

simulate also supports --rpc-url <rpc> with resolution order: --rpc-url -> RPC_URL_<chainId> -> RPC_URL.

When you are ready to execute live:

grimoire cast spells/uniswap-swap-execute.spell --key-env PRIVATE_KEY --rpc-url <rpc>

Want live snapshot params for strategy inputs?

grimoire venue morpho-blue vaults --chain 8453 --asset USDC --min-tvl 5000000 --format spell
grimoire venue pendle chains

Before running venue actions in a new environment, validate adapter/env/RPC readiness:

grimoire venue doctor --chain 1 --adapter uniswap --rpc-url <rpc> --json

If this command errors with Unknown venue adapter "doctor", your installed CLI is outdated. Use one of:

• npm i -g @grimoirelabs/cli@latest
• npx -y @grimoirelabs/cli@latest venue doctor --chain 1 --adapter uniswap --rpc-url <rpc> --json
• bun run packages/cli/src/index.ts venue doctor --chain 1 --adapter uniswap --rpc-url <rpc> --json

Advisory steps (advise) call Pi when a model is configured (spell model, CLI model/provider, or Pi defaults). If no model is available, the runtime uses the spell’s fallback. Record advisory outputs with simulate (or cast --dry-run), then replay deterministically with --advisory-replay for live execution.

Advisory docs:

• docs/how-to/use-advisory-decisions.md
• docs/explanation/advisory-decision-flow.md
• docs/reference/spell-syntax.md#advisory-syntax

Protocol comparison docs:

• docs/how-to/compare-protocol-metrics.md
• docs/reference/venues.md#metric-surfaces

See grimoire --help for all CLI commands.

Library entry point (@grimoirelabs/core)

Use this when embedding Grimoire into an app/agent process. The library shares the same compiler/interpreter semantics and preview/commit model as the CLI.

See docs/reference/compiler-runtime.md for compile, preview, commit, execute, and session APIs.

Agent-assisted entry point (skills)

Use skills in skills/ so agents can work immediately with Grimoire:

• install: npx skills add https://github.com/franalgaba/grimoire
• skills/grimoire/ for install, CLI usage, syntax starter, and runbook
• venue skills for snapshot params (skills/grimoire-aave/, skills/grimoire-uniswap/, skills/grimoire-morpho-blue/, skills/grimoire-across/, skills/grimoire-hyperliquid/, skills/grimoire-pendle/, skills/grimoire-polymarket/)

For Claude Code, run the same install command in the Claude Code terminal and start a new session to load the skills.

---

Example spell

spell YieldOptimizer {

  assets: [USDC, DAI]

  advisors: {
    risk: {
      model: "anthropic:haiku"
    }
  }

  venues: {
    aave_v3: @aave_v3
    morpho_blue: @morpho_blue
  }

  params: {
    amount: 100000
  }

  on hourly: {
    decision = advise risk: "Do gas costs justify rebalancing now?" {
      output: {
        type: boolean
      }
      timeout: 10
      fallback: true
    }