Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Black Duck SCA — MCP Servers
Connect your AI coding assistant to the Black Duck Software Composition Analysis (SCA) products. The MCP servers expose your Black Duck instances as a set of tools that any MCP-compatible AI harness can call to perform actions such as: investigate your software's security posture and vulnerabilities, triage findings, scan code and binaries, and generate vulnerability reports and SBOMs.
Table of Contents
Products
This package ships two MCP servers, one per product. Install the one you have an instance for, or both. Each has its own guide with the installation steps, the full list of tools/resources and environment variables, and troubleshooting.
| Product | MCP Guide | Use it to |
|---|---|---|
| Black Duck SCA | BDSCA | Explore Projects and BOMs, triage Vulnerabilities, check Policy compliance, review component matches, run scans, generate SBOM, VEX/CSAF and Notices reports |
| Black Duck Binary Analysis | BDBA | Scan binaries and firmware, review the components found in them, triage Vulnerabilities, override component versions, generate reports |
You can always check the full list of available tools by asking your AI assistant to list them, e.g. "What Black Duck SCA tools do you have access to and what are their functions?"
Latest changes
See the changelog for the latest changes.
Getting Started
Prerequisites
- A Black Duck instance (SCA, Binary Analysis or both) with a user account and API token
- Python 3.13
- [uv](https://docs.astral.sh/uv/getting-started/installation/) — Python package manager
- Java 11+ _(optional)_ — required only for BDSCA source scanning via Detect
Both servers run from the same blackduck-sca-mcp package, with uvx.
Installation
The guides have the setup instructions for various assistants, e.g. Claude Code, Claude Desktop, VS Code, Copilot CLI and Roo Code:
Once connected, ask your assistant "Am I connected to the Black Duck instance?" to check the setup. If something goes wrong, see the troubleshooting sections of the guides:
Security
Token permissions
API tokens inherit the full permissions of the associated user account. If you provide a token with write access, the AI assistant can modify data in your Black Duck instance — including updating vulnerability remediation status, policy overrides and triages.
We recommend creating a dedicated service account with the minimum permissions required for your use case. See the Role and Permission Matrix for details.
Read-only mode
Set BLACKDUCK_MCP_READ_ONLY=true, or start a server with --read-only, to turn off the tools that change data or write files. The tools it turns off are listed in Read-only Mode for BDSCA and for BDBA.
Token storage
Most AI assistants pass their environment to every MCP server and tool they start, so a token in an env var can be read by all of them. Each server looks for its token in this order:
BLACKDUCK_BDSCA_TOKENorBLACKDUCK_BDBA_TOKEN, the token itselfBLACKDUCK_BDSCA_TOKEN_FILEorBLACKDUCK_BDBA_TOKEN_FILE, the path of a file with the token~/.blackduck/bdsca-tokenor~/.blackduck/bdba-token, read when neither of the two above
is set
To keep tokens out of the environment, save each one in a file that only your own user can read:
mkdir -p ~/.blackduck && chmod 700 ~/.blackduck
printf '%s' "<bdsca-api-token>" > ~/.blackduck/bdsca-token
printf '%s' "<bdba-api-token>" > ~/.blackduck/bdba-token
chmod 600 ~/.blackduck/bdsca-token ~/.blackduck/bdba-tokenand configure only the URLs. The health_check tool shows where the token came from.
Data sent to your LLM provider
MCP tools results go to the LLM that your AI assistant uses. These may hold Vulnerability findings, Project, Version and Component names, Policy settings, and Scan results. The API tokens never go to the LLM.
Black Duck does not control how your LLM provider stores or uses this data, or whether the provider trains models on it. Before you use the servers with sensitive or regulated findings, read your provider's terms and data processing agreement, and use its zero data retention or enterprise options where you need them.
Error details
Unexpected error details, which can hold URLs and file paths, stay in the server log and do not go to the AI assistant. Set BLACKDUCK_MCP_DEV_MODE=true to have errors shown to the LLM too, for example if you are debugging a problem.