AGENT SCAN / INTERACTIVE PREVIEW
Would you attach
this tool?
A tool description can leave out something important. Read three examples, choose a decision, then see what the checker finds.
No account. No install. Your manifest stays in this browser.
01 / READ THE METADATA
What should the static check decide?
These are synthetic teaching examples. Match the scanner's decision on supplied metadata; deployment still needs its own review.
EXAMPLE 1 OF 3
A small calculator
Inspect the complete synthetic JSON
That is your match count for these examples, not an agent rating. Try your own manifest below, or let a teammate try the same challenge.
02 / CHECK YOUR TOOLSET
Keep the file. See the findings.
Paste an MCP tools/list response, or choose a JSON file from your device. This preview checks up to 200 tools in a file of up to 256 KiB. It does not connect to a server or execute tools.
How do I get a tools/list file?
Export the JSON response to tools/list from your configured MCP client or server's developer tools. If the response is paginated, export the complete tool list. Do not paste connection credentials or a server URL here. Other formats and larger manifests can use the local CLI.
Processed in memory. Nothing is uploaded or saved by this page. Browser extensions and device policies are outside this page's control.
No blocking finding means no configured blocking rule fired on this metadata. Runtime permissions, actual execution, and deployment behavior were not assessed.
Take it to a review
The summary below omits tool names, findings, raw input, and fingerprints. Review it before copying it into a team conversation or pull request.
Repeat it in your repository
Run the pinned check on a committed manifest in GitHub Actions. It reports a static metadata decision for that commit, not a certification or runtime result.
Metadata is one part of the decision.
Agent Scan checks supplied tool metadata before attachment. A self-run result is not a BackBond Score or an attestation of runtime behavior.
BackBond is the rating-and-insurance layer for AI agents. Agent Scan is its public tool for examining tool metadata.
Running this package does not create insurance coverage, determine eligibility, or imply that BackBond has verified the environment.