Protecting Smart Contracts 24/7

Catch Issues
Before the Audit. Gate Every PR After.

Multi-model AI plus deterministic rules for Solana, EVM, CosmWasm, and Cosmos SDK. Findings come with exploit sketches, PoCs, and fixes, and the GitHub Action screens every pull request.

Source deleted after scanState-of-the-art models, listed in privacy policyNot used for training

SOLANA · EVM · COSMOS

Solana
EVM
Cosmos SDK
&
CosmWasm

Sui and Move on the roadmap - see the roadmap board

47%
False Positives Filtered, Zero TPs Lost
90%
Severity Accuracy on True Positives
3
AI Models Cross-Validating
~3 min
Typical Deep Scan Time

OdinBench measured on 52 Solana and EVM contracts with known ground truth (100% Solana recall, 77% EVM recall) - methodology and results

WHAT ODIN SCAN DOES

This Does Not Replace a $50k Audit.
It Makes That Audit Cheaper - and Keeps PRs Clean After

A human audit is a point-in-time snapshot - expensive, and stale the moment you merge again. Odin Scan is the scanner you run before the auditors arrive and after they leave: walk in with a clean repo, pay for judgment instead of typo-hunting, then keep every pull request screened.

Multi-Model Cross-Validation

State-of-the-art models analyze the same contract and must agree before a finding ships. We continuously upgrade to the best available models as they are released. Disagreements get flagged, not buried.

Exploit Sketches and PoCs

Every finding explains how an attacker would reach the bug, includes a proof-of-concept where possible, and proposes the fix.

Verification Pass

A dedicated post-scan stage hunts false positives and flags them in your report instead of silently dropping true positives.

SARIF and PR Comments

Findings land as branded PR comments and SARIF uploads to the GitHub Security tab, so reviewers see them where they work.

Merge Gate on Every PR

Wire severity policies into CI: the Action blocks merges when new critical or high findings appear, on every plan including trial.

Solana, EVM, and Cosmos - All First-Class

Dedicated rules for Solana (Anchor, native, Pinocchio), EVM/Solidity, and CosmWasm/Cosmos SDK: storage collisions, unsaved state, unsafe math, signer checks, and more. No platform is an afterthought.

INTEGRATIONS

Fits Into Your Workflow

Connect Odin Scan to the tools your team already uses.

GitHub Actions

Run scans on every pull request with branded PR comments, SARIF uploads to the Security tab, and repository-level configuration.

Slack Notifications

Get instant alerts in your team's Slack channels when critical vulnerabilities are found. Severity-based policies let you control what triggers a notification.

API Access

Trigger scans, fetch reports, and manage repositories programmatically. Integrate Odin Scan into any CI/CD pipeline or custom workflow.

Constantly evolving - new rules, improved detection models, and expanded platform support added regularly.

Odin Scan catches the classes of bugs that rules and LLMs see reliably: missing access control, unsafe math, storage key collisions, input validation, predictable randomness, and more. It does not replace a manual review of novel economic logic. The findings report is your pre-audit cleanup, not a signed human opinion.

SCANNER + CI GATE · NOT AN AUDIT REPORT · SEE AN EXAMPLE FINDING

HOW IT WORKS

See Odin Scan in Action

Watch a quick 2-minute demo of how Odin Scan detects smart contract vulnerabilities with AI-powered precision.

FOR AUDIT FIRMS

Start Every Audit With a
Fast, Thorough First Pass

Odin Scan is your pre-audit triage engine. It finds the obvious issues before your team reads a single line, so you spend your hours on the deep, expensive logic - and quote with confidence.

1. Client connects their repository.

2. We return severity, exploit paths, PoCs, and fixes in minutes.

3. You triage, confirm, and still sign the audit.

Cut scoping time

Walk into an engagement with a machine-generated baseline already covering the obvious classes.

Fewer surprises

Catch the easy wins early and avoid basic findings in your final report.

De-risk fixed-fee quotes

Know the scope before you commit. Price engagements from data, not estimates.

See the Auditor Workflow

ODINBENCH

Measured Against Known Ground Truth,
Not Vanity Counts

Every detection claim on this site traces back to OdinBench: an open corpus of Solana and EVM contracts with documented vulnerabilities, run through the same pipeline customers use. Results are committed to the repository, raw and filtered, including the runs where recall was disappointing.

100%
Solana recall on known ground truth
18/18 true positives found - zero missed on the Sealevel Attacks corpus
77%
EVM recall on known ground truth
23/30 true positives found across DeFiVulnLabs and Damn Vulnerable DeFi
41%
Solana F1 after filtering (26% raw)
The falsification gate removed 29 false positives with zero true positives lost

Rescan these contracts yourself

Solana

Missing Signer Check (Anchor)

Sealevel Attacks - missing signer on withdraw; found by the missing_signer_check rule

Solana

Arbitrary CPI (Anchor)

CPI target program ID supplied by the user without validation (critical)

EVM

Reentrancy

DeFiVulnLabs reentrancy contract with the classic ethers/fallback exploit

Sources live in the odin-scan-bench corpus (Sealevel Attacks + DeFiVulnLabs). Paste any of them into the demo scanner and expect the named finding.

PRICING

Priced per Team,
Not per Seat

One flat price covers your whole team and every repo. The GitHub Action is included on every plan - a 7-day Pro trial, cancel anytime from Settings.

7-day Pro trial: full deep-agent scans on your repos, no card required, cancel in Settings in one click. Your code is deleted right after each scan and never used for training.

Odin Scan does not replace a $50k audit - it makes that audit cheaper by walking the auditors into a clean repo, then keeps every pull request screened after they leave.

Frequently Asked Questions

Everything you need to know about Odin Scan

Which platforms?

Solana, EVM, CosmWasm, and Cosmos SDK (Go). Solana and EVM get the deepest dedicated rules. Sui and Move are on the roadmap, not in the product yet.

How accurate is it?

On the public OdinBench corpus, Solana recall reached 100% (18/18 ground-truth vulnerabilities found) and EVM recall 77% (23/30 found), with the falsification gate removing false positives and never losing a true positive. Full methodology below.

What happens to my code?

Cloned to a temporary directory, scanned, then deleted immediately. Nothing is stored and nothing is used to train models. Details in the privacy policy.

Which AI models?

State-of-the-art models running a structured deep-agent audit, with an in-prompt falsification gate. We constantly upgrade to the best available models as they are released - current providers are named in the privacy policy.

How much does it cost?

7-day Pro trial, no card. Flat per-team pricing from the Starter tier - every plan includes the GitHub Action. Full breakdown on the pricing page.

Does it work in CI?

Yes. The GitHub Action runs on every pull request: PR comments, SARIF uploads, and a severity-based merge gate. Included on every plan, including trials.

What blockchain platforms does Odin Scan support?
How does the AI security analysis work?
What happens to my repository after analysis?
How accurate are the vulnerability findings?
Which AI models do you use?
What pricing plans are available?
Can I integrate Odin Scan into CI/CD?