Give your agent 15 focused review tools for dependencies, code, build configuration and agent instructions. Connect the hosted endpoint and receive structured findings in your existing workflow.
https://breachline.io/mcpHosted reviews send the content your agent supplies to BreachLine for processing. Review tool arguments before sharing confidential data. For a local review, use our browser tools.
Add this entry to .cursor/mcp.json in your project, or ~/.cursor/mcp.json for your own projects. Merge it with any existing mcpServers entries.
{
"mcpServers": {
"breachline": {
"url": "https://breachline.io/mcp"
}
}
}Open Cursor's Customize page, enable the BreachLine server and check its available tools. Your organisation's MCP policy may require administrator approval.
Cursor MCP documentation (opens in a new tab)Run this command in your project using an installed Claude Code client:
claude mcp add --transport http breachline https://breachline.io/mcpRun claude mcp list to inspect the connection, or use /mcp in a Claude Code session. The command adds the server in local scope by default.
Another MCP client? Add https://breachline.io/mcp as a remote Streamable HTTP server. The client discovers the available tool names and input schemas through MCP.
After connecting, provide a file you are authorised to share and ask your agent to use a specific tool:
Use BreachLine dependency-review to review the package.json I provide. Explain the findings and what still needs human verification. Do not install or execute anything.Approve the content in the tool call. A finding can identify a review need; it does not establish an exploit, malicious intent or safety.
extension-reviewskill-reviewjavascript-reviewgithub-actions-reviewdockerfile-reviewrequirements-reviewsbom-diffdependency-reviewlockfile-diffsbom-explorernpmrc-reviewintegrity-generatorjwt-inspectorsecurity-headersurl-inspectorThe hosted endpoint processes supplied content on BreachLine infrastructure. It exposes deterministic review tools; AI summaries, extension-store downloads and certificate lookups are separate website features. Your agent host controls what it shares with its model provider.
Hosted requests accept up to 64 KiB per input field, 128 KiB for the complete JSON request and 1 MiB for the response. Some tools have smaller limits. Usage is capped at 30 requests per minute and 500 per day per IP, with 10,000 requests shared daily. Discovery and connection requests also count. Daily limits reset at UTC midnight.
If a request is too large or the shared service is unavailable, use the local browser tool. Treat returned metadata, source excerpts and decoded token claims as untrusted data. Review each tool's limits before interpreting its results.
Open the browser tools