Vigilance vs SafeDep

What SafeDep Does Well

SafeDep blocks malicious packages at the moment a developer installs them. Its open source tool PMG wraps the package manager and checks every install before any code runs. PMG covers npm, pnpm, yarn and bun for Node.js. It also covers pip, pipx, poetry and uv for Python. A developer does not need to change a habit.

PMG uses three layers. The first layer checks the package against SafeDep's malware database. The second layer is a dependency cooldown. It blocks recently published versions for a set window, so a new compromised release has time to be found. The third layer is an optional sandbox. It uses macOS Seatbelt or Linux Landlock, with Bubblewrap as a fallback, to limit what an install can touch.

The cooldown is a strong idea. Most supply chain attacks are found within hours or days of release. A team that waits a few days before it takes a new version avoids many of them. SafeDep builds that wait into the tool.

SafeDep also ships vet, an open source command line tool for dependency scanning. Vet reads npm, PyPI, Maven, Go, Ruby, Rust and PHP packages. It also reads container images and SBOM files in CycloneDX and SPDX formats. Its policies use CEL expressions. A team can write rules for licenses, vulnerability limits and scorecard needs. SafeDep says vet analyzes how your code uses a dependency, to rank real risks above CVE noise.

The cost of entry is low. PMG and vet use the Apache-2.0 license. PMG needs no account or API key. SafeDep also lists an MCP server for AI coding agents, an Agent API, and an inventory of AI agents and SDKs. Its platform claims SOC 2 Type II and ISO 27001:2013 certification. Those items matter to a buyer with a security questionnaire to answer.

SafeDep reaches many ecosystems. Its site lists npm, PyPI, Go, Maven, RubyGems, Cargo, NuGet and Docker registries. It also connects to GitHub, GitLab, Cursor, Claude and Windsurf.

Where It Falls Short

SafeDep falls short where a threat feed always falls short, which is on an attack that nobody has listed yet. Its protection depends on a verdict from SafeDep Cloud. A package that the feed has not seen has no verdict to apply.

SafeDep does cover that gap in part. The cooldown delays new versions, and the sandbox limits damage. Both have a cost. A cooldown also delays a legitimate security patch. A sandbox limits an install step, but it does not tell you what a new file in the package can do once your own program loads it.

The second limit is the network. PMG checks packages against SafeDep's community API. SafeDep says the tool needs no account. The PMG page also says the threat intelligence layer needs internet access for the best protection. SafeDep does not document an offline mode. A disconnected or air-gapped site cannot rely on it. Vigilance Pro opens no connection at all.

The third limit is the stage where SafeDep acts. PMG and vet act at install time and in pull requests and builds. A program that arrives some other way does not pass through a wrapped package manager. Examples are an installer, a signed desktop app, a container image or a Linux package archive. Vigilance reads inside deb, rpm, npm, pip, containers, MSI and ISO files. It works on those formats from one binary.

The fourth limit is scope. SafeDep focuses on open source packages and on AI coding agents. It does not claim to compare a trusted version of a vendor program with its update. The xz Utils backdoor and similar cases shipped in a release archive. Vigilance targets that shape of attack.

Some of these limits depend on your setup. A team that only installs npm and PyPI packages on laptops with a network will meet fewer of them.

Feature Comparison

The two tools overlap less than their names suggest. SafeDep decides from a feed, a cooldown and a sandbox. Vigilance decides from the files themselves. The table lists each difference that we can source.

A cell that says "Not documented" means we found no statement on the SafeDep pages we read. It does not say the feature is missing.

When SafeDep Is the Better Fit

SafeDep is the better fit when your main risk is a developer or an AI agent that installs a package from npm or PyPI. PMG sits in the install path and stops a listed package before it runs. Vigilance does not block an install.

Choose SafeDep in these cases.

• You want a blocker at install time, with an open source tool and no account.
• You want a cooldown that delays new package versions across a team.
• You want sandboxed installs on macOS or Linux.
• You need vulnerability scanning, SBOM output and policy as code in pull requests. Vet does this.
• You want to inventory the AI agents and SDKs your company uses.
• Your buyer asks for a vendor with SOC 2 Type II and ISO 27001 reports, plus SSO and RBAC.

Choose Vigilance when the risk sits in the update itself. This covers a vendor installer, a signed app, a container image, a Linux archive or a build output. It also covers a site with no internet. Vigilance shows one file and one new capability. A normal update stays quiet.

Many teams will want both. SafeDep stands at the door of the package manager. Vigilance reads what came through. The two checks use different evidence, so one does not replace the other.

Pricing

SafeDep has a free plan, a Team plan at $100 per month for 5 endpoints, and a custom Enterprise plan. Its open source tools cost nothing. We took these numbers from the SafeDep pricing page.

The Free plan costs $0 and covers up to 3 SDLC endpoints. It includes real-time malicious package protection, 7 days of findings history and community support. It allows 10 package scans and 25 repository scans per endpoint per month.

The Team plan costs $100 per month per 5 endpoints, which is $20 per endpoint. It adds 90 days of findings history, an audit trail and email support. SafeDep lists an annual option with a 20% discount. It also offers a 30-day trial with no credit card, and the trial converts to the Free plan. The Enterprise plan adds SSO, RBAC, MDM rollout, SIEM and EDR integrations, and dedicated support. Its price depends on volume. A separate Threat Intel subscription covers the malicious package feed and API access.

SafeDep counts an endpoint as one point where third-party code enters. That can be a developer machine running PMG, a CI pipeline, a repository with pull request scanning, or an AI coding agent. The price grows with the number of those points.

Vigilance prices by company size and never by machine count. The Free plan costs $0 and runs forever on up to 50 machines. It needs a network and it sends telemetry. Pro starts at $999 per month or $9,990 per year, in Canadian dollars, flat for the whole company. Pro works offline and has no phone home. See Vigilance pricing for every band.

The models differ. SafeDep scales with endpoints. Vigilance steps on company size. A small team with few endpoints can stay free on both tools.

Which One Do You Need?

You need SafeDep if you want to stop known bad packages at install time, and Vigilance if you want to see new capability in any update. Many teams need both.

SafeDep answers the question, "Does a feed list this package as malicious?" Vigilance answers a different question. It asks, "Can this file do something the last version cannot do?" The first question works well on common npm and PyPI attacks. The second works on attacks with no listing, and on software that does not arrive through a package manager.

Known Risk versus New Risk