Your AI assistant is writing outdated code.
Vibgrate feeds it version-correct context from your lockfile, gives it a deterministic map of your real code, scores how far your stack has drifted, and flags the CVEs hiding in your stale dependencies — from one CLI.
Code Drift Intelligence — how far your code, and the AI writing it, has drifted from current, correct, well-understood truth. Not infrastructure drift, not ML model drift.
Free CLI · results in about 60 seconds · never reads your source · no signup
A replay of the actual CLI against our test repos — nothing executes in your browser.
The workflow changed. The guardrails didn't.
AI now writes a large, growing share of production code — at machine speed, against "latest", generic, or hallucinated context. It ships code that targets deprecated APIs, pulls vulnerable dependencies, and misreads the structure of the very repo it is editing. Codebases drift from current, supported truth faster than ever, and almost no team can see it happening.
The teams that win the next few years are the ones whose humans and AI work from the same current, verifiable ground truth.
Three kinds of drift. One system that closes all three.
Give your AI the context and the map to stop guessing — and measure the dependency drift that proves the risk is real, not vibes.
Your AI writes against "latest" — and hallucinates deprecated APIs.
Vibgrate AI Context serves library docs pinned to the exact versions in your lockfile — plus your code map and offline drift — to your AI assistant. Offline, no account.
Version-correct contextAgents re-read your repo every session, burning tokens.
Vibgrate Graph gives them a deterministic, committable map of your real code — so they answer in one tight call instead of re-deriving what they could have been told.
A map your AI can trustRuntimes fall behind LTS; dependencies rot silently.
Vibgrate scores how far you have drifted, 0–100, and turns it into an upgrade plan you ship. Add --vulns to surface the CVEs hiding in your stale dependencies.
DriftScore, explainedMeasure it. Prioritize it. Ship it.
A 0–100 DriftScore for every project, rolled up to repo, team and portfolio.
How the DriftScore worksA prioritized upgrade plan — effort, risk and cost-of-delay — not a wall of alerts.
Tour Vibgrate CloudHand remediation to the AI agent you already pay for. Vibgrate verifies the result.
AI integrations & MCPThe overview your team shares
The portfolio view Vibgrate Cloud opens with — DriftScore, critical risks, and per-project health across every repo. Running here on example data.
Real dashboard widgets · swipe to explore
Example data · no account, no card · push one scan to replace it with your own
Five products. One drift problem.
Free tools for the developer. A shared dashboard for the team. Each piece works on its own — together they cover context, map, and measurement.
Vibgrate CLI
FreeOne command. A 0–100 DriftScore in about 60 seconds.
- Scans ~19 ecosystems — Node, .NET, Python, Java, Go, Rust, Docker, Terraform and more
- Add --vulns to check every dependency against known advisories, with fix versions
- SARIF output and CI gates: --fail-on, --drift-budget, baselines
Vibgrate Graph
FreeA map of your codebase — and what breaks if you change it.
- vg impact shows affected call sites and the tests that cover them
- vg hubs, areas and oddities reveal structure and surprises
- Committable and team-shareable with vg share — readable by any AI
vg impact maps a changeThe green path is what a change touches — and the tests that cover it.
Vibgrate AI Context
Free · offlineStop your AI guessing which version you run.
- Library docs pinned to the exact versions in your lockfile — not "latest"
- Served together with your code map and offline drift, over local MCP
- No account, nothing uploaded — vg serve and it works
Vibgrate Cloud
Free plan availableThe dashboard your whole team works from.
- DriftScore trends, dependency and vulnerability registers, audit-ready reports
- Policies, drift budgets, approvals and an audit trail on the higher plans
- Unlimited seats on every plan — you pay per project, not per person
Vibgrate MCP
With Vibgrate CloudAsk your AI assistant about your whole estate.
- Hosted MCP server over your Vibgrate Cloud workspace data — distinct from the local, offline Vibgrate AI Context
- DriftScores, vulnerabilities, dependencies and approvals — answered in Claude, Cursor or ChatGPT
- One-click OAuth, read-scoped tokens, tenant-isolated
Built for every layer of the team
Developers
Feed your AI version-correct context, and see what a dependency upgrade breaks before you open the PR.
Tech Leads
Per-repo health, architecture erosion and the upgrades worth doing next — from a tool that never reads your source.
Platform Engineers
Drift and vulnerability gates in CI — SARIF out, meaningful exit codes, fully offline.
Security Engineers
Cut CVE noise by fixing the staleness that generates it. SBOMs and CRA-aligned remediation metrics.
Engineering Managers
See which teams are drifting and what one sprint of upgrades buys you.
CTOs
Is your estate modernizing or drifting? Portfolio risk and AI-debt on one board-ready page.
We don't sell you a destination.
Hyperscaler tools modernize you toward their cloud. Vibgrate is independent: we measure, prioritize and verify — and route execution to whatever agents and platforms you choose.
Pay for what drifts.
Scan everything free. Pay per project under governance — not per developer, not per seat. Tiny projects bill as fractions: a nano project counts 1/25, a micro project 1/10. Unlimited seats on every plan.
Get your DriftScore in minutes.
Install the CLI, run one scan, see exactly how far behind your estate has drifted — and hand your AI the context to stop writing outdated code.
# Run without installing anything
npx @vibgrate/cli scan
# Pin a version or see every command
npx @vibgrate/cli@latest --helpno install·Nothing is installed globally — ideal for CI or a one-off scan.
No sign-up required · results in under 60 seconds · never reads your source