atlas-mcp

Both the provided material and objective checks indicate no required secrets or environment variables. No direct exposure of API tokens, cloud credentials, or account secrets is evident; credential handling does not appear to be the primary risk here.

Although no remote endpoint is declared in the manifest, the description says it will “connect ... to Atlas knowledge hub for context retrieval and work capture,” which functionally implies possible outbound communication and data upload. Because the README is absent and endpoints are undisclosed, it is unclear whether user context is sent externally, so this warrants caution rather than a confirmed high-risk rating.

The objective checks explicitly indicate that this tool executes code. For an MCP tool, spawning local processes or executing code is a standard capability, but it still means the runtime environment should be constrained and any system commands or subprocess behavior should be reviewed.

“Context retrieval” and “work capture” suggest the tool may read project context and record work outputs, but the material does not specify which files, directories, or resources are accessed, nor whether it writes local data. The access scope is under-documented, so it should be evaluated under a least-privilege assumption in an isolated environment.

A positive factor is that there is an auditable open-source repository, which materially lowers the risk. Points to watch are that it comes from a third-party registry, has no declared license, shows 0 stars, and has unknown maintenance status; these indicate weaker supply-chain maturity and continuity signals, but not enough on their own to justify a high-risk rating.