OSS Audit MCP
Supports auditing npm packages with license and copyleft classification, release recency, dependents, repository/OpenSSF Scorecard data, and OSV vulnerabilities via package reports, batch license checks, vulnerability lookups, and package comparisons.
Supports auditing NuGet packages with license and copyleft classification, release recency, available repository/OpenSSF Scorecard data, and OSV vulnerabilities via package reports, batch license checks, vulnerability lookups, and package comparisons.
Supports auditing PyPI/Python packages with license and copyleft classification, release recency, dependents, repository/OpenSSF Scorecard data, and OSV vulnerabilities via package reports, batch license checks, vulnerability lookups, and package comparisons.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
OSS Audit MCP server (open-source package due diligence)
An MCP server that lets AI agents vet open-source dependencies before adoption, in a licence review, or during technical due diligence. For npm, PyPI, Go, Maven, Cargo (crates.io) and NuGet it reports licences with copyleft classification, release recency, dependents count, the linked repository's OpenSSF Scorecard, and known vulnerabilities with severity and fixed versions. It reads deps.dev and OSV.dev live and needs no API key.
Not legal advice and not a full security audit. Licence data is detected from package metadata and can be incomplete or wrong (e.g.
non-standard). A clean vulnerability result only means OSV has no advisory, not that the code is safe.
Tools
Tool | What it does |
| One package version: licence + copyleft category, latest version, release recency, releases in the last 12 months, dependents, repo stars/forks/issues, OpenSSF Scorecard (overall + checks), OSV vulnerabilities, red flags |
| Licence table for up to 50 packages, with a copyleft flag (GPL/AGPL/LGPL/MPL/EPL and similar), category counts and per-row errors |
| OSV advisories for a package, or for one exact version. Aliases are merged, items are sorted by severity, and each has fixed versions |
| Side-by-side health table for up to 10 alternatives in one ecosystem, plus the best package per metric |
Example calls
// package_report
{"ecosystem": "npm", "name": "express"}
// -> 5.3.0, MIT (permissive), 115k dependents, repo 69.7k stars, Scorecard 8.5, 0 vulns, flags []
{"ecosystem": "maven", "name": "org.apache.logging.log4j:log4j-core", "version": "2.14.1"}
// -> 8 advisories incl. 2 CRITICAL (Log4Shell), flag "not the default/latest version (2.25.5)"
{"ecosystem": "go", "name": "github.com/gin-gonic/gin"}
// check_licenses (strings "ecosystem:name@version", version optional; or objects)
{"packages": ["npm:express@4.21.2", "pypi:requests", "maven:org.hibernate:hibernate-core@5.6.15.Final",
"go:github.com/gin-gonic/gin@v1.9.1", "cargo:serde", "nuget:Newtonsoft.Json@13.0.3",
{"ecosystem": "pypi", "name": "PyQt5"}]}
// -> hibernate-core LGPL-2.1-or-later (weak-copyleft), PyQt5 GPL-3.0 (strong-copyleft), the rest permissive
// vulnerabilities
{"ecosystem": "npm", "name": "lodash", "version": "4.17.15"} // 5 advisories: 3 HIGH, 2 MEDIUM
{"ecosystem": "pypi", "name": "django", "limit": 5, "include_details": true} // all 160+ ever filed, top 5
// compare_packages
{"ecosystem": "pypi", "names": ["requests", "httpx", "aiohttp", "urllib3@1.26.0"]}
{"ecosystem": "npm", "names": ["axios", "node-fetch", "got"]}Package naming: npm express, @types/node; PyPI requests (case-insensitive); Go module path
github.com/gin-gonic/gin; Maven groupId:artifactId; Cargo serde; NuGet Newtonsoft.Json. The ecosystem accepts
aliases such as python, rust, crates.io, golang and dotnet.
Licence categories
permissive · weak-copyleft (LGPL, MPL, EPL, CDDL, …) · strong-copyleft (GPL, EUPL, OSL, CC-BY-SA, …) ·
network-copyleft (AGPL, SSPL) · other/proprietary (LicenseRef, BUSL, Elastic, CC-BY-NC) · unknown.
An A OR B expression takes the least restrictive option and sets permissiveAlternative when a copyleft option is
also offered. A AND B takes the most restrictive. copyleftFamilies lists GPL/AGPL/LGPL/MPL/EPL hits explicitly.
Licence exceptions (WITH Classpath-exception-2.0) are not weighed, so review those by hand.
Severity
The GitHub advisory rating is used where OSV has one (MODERATE is shown as MEDIUM). Otherwise the server computes
the CVSS v3.x base score from the vector. CVSS v4-only advisories without a rating show UNKNOWN. Advisories that
alias each other (GHSA / PYSEC / CVE / GO / RUSTSEC) are merged into one entry, which keeps the GHSA record and lists
the others in alsoKnownAs.
Related MCP server: closeread-verify
Limits and data caveats
"Latest version" is the registry default. When that default is a pre-release (e.g.
1.0.0.dev6), the newest stable release is used instead.latestReleaseAtis the newest publish of any version, including pre-releases. Go pseudo-versions (untagged commits) are excluded from version counts.Dependents come from deps.dev's v3alpha
:dependentsendpoint. It can change, and it returns 0 or nothing for some ecosystems/packages (e.g. Cargoserdeshows 0; Go and NuGet return none). A note is added when the count is 0.Repository, stars and Scorecard depend on deps.dev linking a source repo. Older versions often lack the link, so the latest version's link is used as a fallback. Scorecard results are weekly snapshots.
vulnerabilitieswithout a version follows up to 3 OSV result pages.truncatedUpstreamis set if more exist.Caps: 50 packages per
check_licensescall, 10 percompare_packagescall, 100 advisories returned, 6 concurrent upstream requests, 30 s HTTP timeout, 120 s per tool call, 12 MB per upstream response.
Data sources and attribution
deps.dev API v3 (Open Source Insights, Google),
https://api.deps.dev. Its generated data is licensed CC-BY 4.0, so every response carries anattributionfield. Keep it when you redistribute results. Attribution: "Package, licence, dependents and Scorecard data from deps.dev (Open Source Insights by Google), licensed CC-BY 4.0."OSV.dev API v1,
https://api.osv.dev. Advisory records are aggregated from upstream databases (GitHub Advisory Database, PyPA, Go, RustSec, …) under those sources' own licences. I did not check OSV's terms during this build. Check them before commercial use.The User-Agent is
oss-audit-mcp/0.1 (+MCP server; ...).
Run
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
.venv/bin/python src/server.py # streamable-http on 0.0.0.0:8080, endpoint /mcp
PORT=8097 .venv/bin/python src/server.py # other port
MCP_TRANSPORT=stdio .venv/bin/python src/server.py # stdio, for local MCP clients
.venv/bin/python tests/live_test.py # live smoke test of every toolEnvironment variables: HOST (default 0.0.0.0), PORT (default 8080), MCP_TRANSPORT (streamable-http, stdio or sse; default streamable-http). No API key is needed.
mcpize.yaml describes the build and start commands for MCPize.
Use with Claude Desktop or another MCP client
Local (stdio). Clone the repo, create the venv as above, then add this to claude_desktop_config.json
(or your client's MCP config), with the paths changed to where you cloned it:
{
"mcpServers": {
"oss-audit": {
"command": "/path/to/oss-audit-mcp/.venv/bin/python",
"args": ["/path/to/oss-audit-mcp/src/server.py"],
"env": {"MCP_TRANSPORT": "stdio"}
}
}
}Remote (streamable HTTP). Run .venv/bin/python src/server.py on a server and point any client that supports
remote MCP servers at http://<host>:8080/mcp. GET /health returns {"status": "healthy"}.
License
MIT, see LICENSE. The data returned by the tools stays under the terms of its sources (see above).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Package verdicts for AI coding agents: check npm, PyPI, Go and NuGet installs before they run.
Check an npm/PyPI package before an AI agent installs it: 5 focused supply-chain tools.
Verify npm packages before your AI agent installs them: hallucinations, advisories, API drift.
Package intelligence for AI agents across npm, PyPI, crates.io and deps.dev. No API keys.
Related MCP Servers
- AlicenseAqualityCmaintenanceChecks npm and PyPI packages for outdated versions, deprecation status, and breaking changes with cited sources, enabling AI agents to verify dependency freshness.16 npmISC
- AlicenseAqualityDmaintenanceProvides verified dependency-audit verdicts for AI agents, checking installed versions against OSV advisories and splitting direct vs transitive dependencies.3MIT
- AlicenseNot gradedqualityFmaintenanceOpen source intelligence for AI agents — GitHub project-health scoring, package dependency-risk analysis, trending repositories, license checks, and side-by-side package comparison.MIT
- AlicenseAqualityDmaintenanceDependency security & health auditing for AI agents with no account or API key required.22MIT