@cyanheads/osv-advisory-mcp-server
Query OSV.dev for package vulnerabilities, batch-audit dependency lists, and fetch full advisory records via MCP. STDIO or Streamable HTTP.
5 Tools · 1 Skill
Overview
Vulnerability data from OSV.dev, the open-source vulnerability database. Query a package version or Git commit, batch-audit a dependency list or SBOM, and fetch complete advisory records with CVSS severity, CVE aliases, and affected version ranges.
Tools
Capability reference
osv_query_package tool
- Provide
name, ecosystem, and version, or a purl such as pkg:pypi/[email protected]. Known ecosystem bases normalize case (pypi → PyPI); release and repository suffixes retain their spelling. A PURL needs exactly one version: embedded in the URL or supplied separately. Versions are exact, never ranges
- Alpine needs a release branch, such as
Alpine:v3.17 or distro=alpine-3.17.3 in a PURL. Debian PURLs accept known releases or codenames, with an optional debian- prefix. Numeric Ubuntu PURL qualifiers such as distro=ubuntu-22.04.1 select the release Ubuntu:22.04
- For a release-qualified Ubuntu query, supply the installed binary name and version. Both package tools look up the source packages OSV indexes for that release. For example,
libssl3 on Ubuntu:22.04:LTS resolves to openssl. A fallback lookup covers only the supplied name and marks coverage incomplete. A mapping rate limit or exhausted deadline requires a retry
- Maven needs
groupId:artifactId (a single groupId/artifactId slash normalizes). Composer PURLs map to Packagist; an optional repository_url keeps repository scope
- Unsupported ecosystem-specific constraints fail as
invalid_version; npm/crates.io exact pins with = or == normalize to the installed version
- Each advisory carries CVE
aliases, optional upstream IDs, severityLabel with its severitySource, the queried package's fixedVersions, affectedRanges, and cweIds
truncated: true indicates more OSV pages remain. Pass nextCursor as cursor with the same identifier to continue. Repeat that input cursor alongside page when retrieving fields from a resumed reply. An empty initial result confirms no known vulnerabilities when lookup and source resolution completed. An empty terminal resume only confirms exhaustion of the continuation
- When upgrading from before 0.6.0, restart any Ubuntu PURL query with a numeric distro qualifier by omitting its old cursor. The new query scopes findings to the requested release
osv_query_batch tool
- Submit 1–5,000
{name, ecosystem, version} tuples or {purl, version?} entries. The call processes the first 1,000. Set OSV_BATCH_SERVE_OVER_CAP=false to reject submissions above 1,000
summary.submittedPackages, unprocessedCount, and nextInputOffset account for the input. If nextInputOffset is set, submit packages.slice(nextInputOffset) in a new call and add that offset to its row indices. All outcome counts describe the processed prefix
- Each row carries
vulnerable, vulnCount, vulns, truncated, and error; a failure lands in the affected rows' error instead of failing the call
summary totals the scan (vulnerableCount, cleanCount, truncatedCount, errorCount, totalVulns, worstSeverity), and truncated rows never count as clean. The whole scan has a 50-second deadline
vulnerableRows, errorRows, and truncatedRows in summary retain original zero-based indices even when the preview omits those rows. Lists can overlap; retrieve a row with /results/<index>
summary.sourceResolutionUnavailableRowRanges identifies rows checked under their supplied Ubuntu names after a mapping failure. For example, 0,3-5 means rows 0, 3, 4, and 5. Those rows are excluded from cleanCount
- Resume a row through
osv_query_package with its original identifier and nextCursor as cursor
- If a nested advisory page returns a timeout or upstream error, retry the same path. You can also retrieve
/results/<index>/error for the row diagnostic
osv_query_commit tool
- One full
commit: 40 hexadecimal characters for SHA-1 or 64 for SHA-256; either letter case is accepted, abbreviations are rejected
- Advisories retain record-wide severity, aliases, upstream IDs, and GIT ranges with repository URLs and ordered events. Readable advisory sections retain complex values in fenced JSON. No package-specific
fixedVersions are inferred
- Pass
nextCursor as cursor with the same commit to continue a truncated result. An empty terminal resume confirms the continuation is exhausted
osv_get_vulnerability tool
- One exact advisory ID from any OSV source (
GHSA-, PYSEC-, RUSTSEC-, GO-, DSA-, USN-, CVE-, …), matched case-sensitively
- Full record:
details, aliases, optional upstream and related IDs, every affected package with its ranges and source metadata, severityLabel, cweIds, and references
- When OSV publishes evidence for a converted range, inspect
/affected/<index>/ranges/<index>/database_specific. Its source tags and release boundaries accompany OSV's normalized events
- An unknown ID fails as
vulnerability_not_found; withdrawn is set only on retracted advisories
osv_list_ecosystems tool
- No input; returns the
ecosystems strings the query tools accept, plus a note with the date the list was last checked against OSV.dev
- Returns canonical spellings. Query tools normalize known base names case-insensitively and forward unknown names to OSV.dev; release and repository suffixes remain unchanged
Features
Built on @cyanheads/mcp-ts-core: stdio and Streamable HTTP transports, pluggable auth (none / jwt / oauth), swappable storage (in-memory, filesystem, Supabase, Cloudflare KV/R2/D1), structured logging with optional OpenTelemetry tracing.
OSV-specific:
- No API key required — OSV.dev is public and keyless, with no published rate limit
osv_query_batch finds matching advisory IDs through OSV's /v1/querybatch, then fetches each unique advisory once, so batch rows carry the full records (aliases included) the batch endpoint omits; requests are capped by OSV_BATCH_CONCURRENCY
osv_list_ecosystems is checked against the live OSV schema and API with bun run check:ecosystems
Agent-friendly output:
- Truncation is never read as clean —
truncated flags incomplete OSV pagination, and truncated batch rows are left out of cleanCount
- CVE-shaped
aliases and upstream IDs chain to nist-nvd-mcp-server for CVSS base scores and CISA KEV status. The lists remain separate; a distribution advisory can cover several upstream vulnerabilities
- Response fields are bounded to 24,000 bytes; upstream cursors are returned intact outside that budget. Start retrieval by repeating the original input with
page: overflow.page. Continue the same path with offset: page.nextOffset and revision: page.revision until nextOffset is null. Fetch each entries[].deferred.path separately at offset zero without a revision; never copy the whole output page object into the input
- Advisory text in
content[] escapes HTML and unsafe-link constructs while preserving HTTP(S) links; caller coordinates render literally. Commit data and affected metadata use reversible JSON where literal rendering needs it. structuredContent keeps every OSV string verbatim
Getting started
Use the public hosted instance with Claude Code:
Or run locally through npm:
For other MCP clients, add this to the client configuration:
No API key is required. For Docker or a source build, see Running the server.
Configuration
All configuration is validated at startup, and every variable is optional.
See .env.example for the full list of optional overrides.
Running the server
Local development
Requires Bun v1.4.4 or higher (or Node.js v24+).
-
Build and run:
-
Run checks and tests:
Docker
The image defaults to HTTP transport, stateless session mode, and logs to /var/log/osv-advisory-mcp-server. For stdio, use docker run -i --rm -e MCP_TRANSPORT_TYPE=stdio ghcr.io/cyanheads/osv-advisory-mcp-server:latest. Build locally with docker build -t osv-advisory-mcp-server .; OpenTelemetry peers are installed by default — use --build-arg OTEL_ENABLED=false to omit them.
Project structure
Development guide
See CLAUDE.md/AGENTS.md for development guidelines and architectural rules. The short version:
- Handlers throw, framework catches — no
try/catch in tool logic
- Use
ctx.log for request-scoped logging, ctx.enrich for response context, and ctx.signal for cancellable OSV requests
- Register new tools via the barrel in
src/mcp-server/tools/definitions/index.ts
- Wrap external API calls: validate raw → normalize to domain type → return output schema; never fabricate missing fields
Contributing
Issues are welcome. Run checks and tests before submitting:
License
Apache-2.0 — see LICENSE for details.