Build Solana projects the way agents actually work.
Geppetto turns commands like new, init, test, audit, and deploy into stable workflow contracts that developers and AI agents can execute reliably.
# one repo-local contract
bun ./bin/geppetto-cli.ts deploy --output json --write-back
{
"status": "success",
"cluster": "devnet",
"program_id": "...",
"steps": ["buildProgram", "deployProgram"]
}Not just deployment. A workflow contract.
Most Solana repos hide critical process in shell history, local state, and stale notes. Geppetto makes the repo legible: config, examples, checks, outputs, and agent instructions all line up.
Agent-readable structure
Canonical agent entry files, examples, and knowledge modules tell an agent where the source of truth lives.
Executable primitives
new, init, test, audit, and deploy give the workflow stable command surfaces.
Structured outputs
Deploy results are written as machine-readable artifacts so humans and agents can continue from the same state.
Workflow surface
Deploy is the first productionized path, but the product shape is broader: agent-native Solana engineering workflows.
geppetto new
Create a minimal Pinocchio + Geppetto program scaffold with canonical agent guidance.
geppetto init
Add the repository guidance files that make an existing project easier for agents to continue.
geppetto test
Run the agreed validation surface through a stable local command contract.
geppetto audit
Run format, compile, and optional clippy gates as the baseline review surface.
geppetto deploy
Build, deploy, emit artifacts, and optionally write back the deployed program_id.
Next: review reports
Extend audit from a static gate into structured PR findings and anti-pattern checks.
Already real today
The current repository ships the CLI, SDK knowledge surface, deploy pipeline, mock E2E coverage, release checks, and this Cloudflare-ready site.