Yes, within limits. You add Safeguard to ChatGPT as a custom MCP server, sign in with OAuth, and ask questions about packages, vulnerabilities, SBOMs and license policy in your tenant. The docs do not list a dedicated "check before install" tool, so the dependable flow is checking what is already in your projects.
We started this post as a pre-install check walkthrough. The documentation supports the connection and a set of query tools, so that is what this covers. For a package you have never used, there are other Safeguard surfaces built for that, noted below.
What does the connection give you?
Safeguard hosts an MCP server at https://mcp.safeguard.sh/mcp. Per the docs, once connected ChatGPT can query vulnerabilities, SBOMs, packages, licenses, findings and compliance data, generate remediation plans for npm, pip, Maven, Go and Cargo projects, compare SBOMs in CycloneDX and SPDX, and run policy gate checks. Tool calls over MCP are unlimited on every plan.
One caution from the docs matters for security: nothing needs installing. Safeguard does not publish an npm package or a local bridge for its MCP server, so any package claiming to be the Safeguard MCP server should be treated as untrusted.
How do you connect it?
- Create or sign in to your account at app.safeguard.sh.
- In ChatGPT, open Plugins, select the plus button, then Add custom MCP server.
- Enter a name, for example Safeguard Security, and a description.
- Under Connection, enter the server URL above.
- Choose OAuth. ChatGPT registers itself with Safeguard, so there is no client ID or key to copy. Read the risk warning, select I understand and want to continue, then Create as a plugin.
- When ChatGPT sends you to Safeguard, sign in, check the app name and the host that will receive the authorization code, then click Authorize access.
If the option is missing, the docs say to ask your ChatGPT workspace administrator, since workspace policy controls custom MCP servers.
Which tools will it see?
Fewer than you may expect on day one. The docs say a new workspace starts with a small read-only set, and an administrator turns on the rest per tool. That is a good default. Leave write actions off until you have decided who may trigger them. Tenant admins can also switch off any client type, including ChatGPT.
What can you ask about packages?
The docs list prompts such as "Find critical vulnerabilities," "Show me all SBOMs," "Which packages violate my license policy?" and "Generate a remediation plan for my npm project." Phrase package questions in the same way, tied to your projects:
- Which of my projects use this package, and at which versions?
- Are any versions of it in my SBOMs flagged as vulnerable?
- Does it violate my license policy?
Because the assistant acts as you, it sees only what your Safeguard role allows, within the tools your workspace enabled.
What about a package you have not adopted yet?
That is a different question, and another surface fits it. The free in-browser scan at gold.safeguard.sh/scan checks a manifest of up to 100 dependencies, and the Gold Open Source Chrome extension does package lookups in the browser. Use those before you add the dependency, and use the MCP connection to track it once it is in a project.
Try the MCP server setup demo to see the flow.
What are the limits?
An assistant can summarize and look things up. It should not be the only gate before an install. Use CI policy gates for enforcement, and keep a human in the loop for any change the assistant proposes. Treat answers as a starting point and confirm the finding in the Safeguard app before you act on it.
What to do
- Connect ChatGPT with OAuth, not a shared key.
- Keep the default read-only tools until you have a reason to enable more.
- Enable write tools one at a time, and tell your team which ones.
- Ask package questions about your own projects first.
- Use the free scan or the Gold extension for packages you are evaluating.
- Put a CI gate on dependency changes so the check does not rely on someone remembering to ask.
Quick check
Do I need to install a Safeguard package for ChatGPT?
No. The server is hosted, and Safeguard publishes no npm package or local bridge for it.
Why can ChatGPT only see a few tools at first?
A new workspace starts with a small read-only set, and administrators enable other tools per tool.
Sources
Last reviewed: 18 October 2026.
