skip to content
// midkernel

AI-first security_

Attackers already use AI. We use it on defense.

Midkernel is an AI-first security solution. Connect a repository, run an open workflow in an isolated sandbox, read the report and the complete log. Pay per run, not per seat.

Read-only GitHub app · credits per run · packs never expire

$ midkernel scan ./repo
# proves what your code actually does
→ report · complete log
// problem

Most scanners guess. Attackers do not.

You get findings, not the prompt, the tools, the model, or where it stopped. Attackers point agents at a known class and repeat it. Defenders should do the same, on purpose.

Scan

Scan a codebase. Get a report: path, root cause, proof, severity with the reason, and the trace. Workflows are open source. Scores live on a public bench. You pick low, balanced, or max. The credit price is on screen before the run starts.

Threat Intel

Ranked threat classes (for example evm.reentrancy), Web2 and Web3, from public sources we ingest and rank. Status is watching, trending, or active-exploit. Insights and a dashboard. Intel ranks classes; it does not invent a fourth Scan profile.

// profiles
profilewhenrun
lowpull requestsFast and cheap. A first pass.
balancedroutine: at least monthly, and when models releaseThe default. Leading models, verification.
maxbefore launches and major versions, or when Threat Intel is trending or active-exploitStrongest models, repeated passes, verification.
// bench

Scores live on a public bench.

// contact

Run one on your repo.