Basenull AI Ops · Supply Chain

See inside any MCP server.

MCP servers are the third-party tools your AI agents reach for at runtime. mcp-inspect surfaces every tool, resource, and prompt a server exposes — so platform, security, and IT teams can review the surface before agents call it in production.

HTTPS endpoint, public servers only. Result is shareable at /i/<id>.

Inspect a known server

What this is for

Pre-flight review

Before letting an agent call a third-party MCP, see the full tool surface — names, descriptions, input schemas, sample JSON-RPC payloads. No CLI install, no copy-paste from someone else's gist.

Shareable inventory

Each inspection has a stable URL. Drop it into a security review ticket, an architecture doc, or a Slack thread when explaining what a server actually does.

Operational triage

When an MCP server changes — tools added, removed, schemas updated — your agents' behavior changes silently. Continuous monitoring is on the roadmap as mcp-watch.

Stdio servers (uvx, npx)

Most published MCP servers ship as uvx or npx commands. Sign in and paste the command — we run it in a basenull-managed sandbox and surface the same tool inventory. Authenticated HTTP MCPs are next.

mcp-inspect is part of Basenull AI Ops: governance, security, observability, workforce enablement, and agent operations for organizations running AI in production. The Model Context Protocol is the open standard agents use to discover and call third-party tools; mcp-inspect is how your team sees what those tools actually do.