One evidence model. Every agent surface.

All five integrations call the same bounded scanner and preserve the same collection status, byte coverage, score version, provenance, and missing-evidence states.

See the value in one request

Ask naturally. Get an audit you can act on.

The integration handles the mechanics. You stay in your current tool, ask for the outcome, and receive evidence—not a black-box grade.

In ChatGPT

“Audit my site for WebMCP readiness and show me the three fixes with the highest impact.”

Evidence, risks, score breakdown, and fixes

Ask where you already work

Paste a public URL or a tool contract into ChatGPT, Claude, Cursor, VS Code, or Chrome.

Call one audit engine

The integration sends the target to the same bounded scanner and versioned evidence model.

Inspect observable evidence

isWebMCP checks discovery, schemas, annotations, safety signals, byte coverage, and evidence gaps.

Fix what matters

Get a category breakdown, prioritized remediation, and a canonical report your team can share.

VS Code extension

Audit a typed or selected public URL and inspect the source evidence in a locked-down editor panel.

VSIX · VS Code 1.96+

Download VSIX

ChatGPT plugin

A remote MCP server plus interactive audit component, explicit tool annotations, and review-ready test artifacts.

MCP endpoint · review kit

Download review kit

Claude plugin

Claude Code plugin with remote MCP configuration and a skill that preserves evidence provenance.

Claude plugin bundle

Download plugin

Cursor plugin

Portable Agent Plugin with the isWebMCP MCP server and an audit skill for implementation work.

Agent Plugin bundle

Download plugin

Chrome extension

Audit the active page in a native side panel without injecting scripts or requesting access to every site.

Manifest V3 · load unpacked

Download extension

Shared remote MCP

Connect without a bundle

MCP clients that support Streamable HTTP can connect directly to the production endpoint. It offers public-source audits, a synthetic walkthrough, imported contract review, and evidence-level guidance.

https://iswebmcp.com/mcp

Honest by construction

The unknowns stay visible

A public-source result cannot prove a tool succeeds at runtime. A contract review cannot prove measured lift. Every integration keeps those boundaries in the interface and distinguishes scanner trouble from website quality.

Inspect the scoring model