Bring your code.
Paste a program, upload a Rust file, or explore a public GitHub repository. Your first review starts right in your browser.
Deep in your codebase. Native to your ecosystem.
Catch the bugs hiding between the lines,
before they ever touch the chain.
Made for the ecosystem.
Right at home in your stack.
A missing signer. An unchecked account. A tiny change with a not-so-tiny consequence. See what a Solana-aware review can surface.
Your Solana program doesn’t live in isolation.
Neither should your code review.
Paste a program, upload a Rust file, or explore a public GitHub repository. Your first review starts right in your browser.
Run Solana-specific checks for common risky patterns. Connect your own AI provider for deeper, contextual analysis.
Get line-level explanations and practical fixes. Export your findings, make the change, and review again.
Good tooling doesn’t replace good engineers.
It gives them the room to do their best work.
The right questions. The right context.
Made for a different kind of runtime.
Spot missing signer checks, unchecked accounts, and unsafe arithmetic before a small oversight becomes an expensive lesson.
pub authority: Signer<'info>✓#[account(has_one = authority)]✓amount.checked_add(deposit)?✓Add project context and review priorities. Pass them into your connected AI review, not into another forgotten document.
# Review priorities
Check PDA seed validation.
Prefer checked arithmetic.
Explain the exploit path.
Browser checks stay in your browser. No wallet connection, no hidden uploads. AI analysis only runs when you explicitly request it.
GOOD THINGS TAKE ROOT.From the first Anchor program to the protocol you can’t stop thinking about. Spend less time chasing the obvious and more time building something that matters.
A useful first step, without a sales call.
No invented limits. No surprise subscription.
A fresh pair of eyes on your next program.
Bring a provider. Keep control of the infrastructure.
The playground checks Rust and Anchor source for common risk patterns: missing signer constraints, unchecked accounts, panic-prone calls, unsafe blocks, and unchecked arithmetic. These are heuristic checks, not a full security audit. An optional AI connection can analyze the submitted source with your project context.
No. Diffodile is a development aid. Findings may include false positives, and a clean result does not mean a program is secure. Use tests, independent audits, and a thorough manual review before deploying contracts that hold value.
Not for local checks. Your workspace stays in this browser’s local storage. Browsing a public GitHub repository contacts GitHub. If you explicitly select AI analysis, your source and project context are sent to this site’s backend and its configured AI provider.
This version supports public repositories without authentication. For private code, upload or paste a file locally. Automated pull-request reviews and private GitHub App access require an additional GitHub integration; they are not enabled in this deployment.
No. Diffodile never asks for your wallet, recovery phrase, or a transaction signature. You can start with the sample program, your own source file, or a public repository.
Set your provider credentials and review access key on the PHP server, then select “Connected AI” in the playground. The deployment guide includes the exact settings. Credentials never belong in the browser or a public JavaScript file.
No credit card. No wallet. Just a little curiosity.