3,400+ MCP servers ready to use
Vinkius

MCP
Inspector

Free compliance diagnostic for the Model Context Protocol specification.

Test and grade any MCP server against the official specification. 34 automated checks across 7 categories. Instant A+ to F compliance grade with step-by-step remediation.

MCP
34 checks 7 categories <15s Free forever
MCP

What is the
MCP Inspector?

The definitive compliance diagnostic for Model Context Protocol servers.

A free, browser-based diagnostic that tests any Model Context Protocol server against the official specification. Paste a URL, get an instant A+ to F compliance grade — across 34 automated checks in 7 categories.

Transport connectivity · JSON-RPC 2.0 protocol validation · OAuth 2.1 authorization · TLS and security hardening · Tool schema quality · Trust and safety scanning · Performance benchmarks — every failed check includes step-by-step remediation. No download, no CLI, no sign-up.

Why test your MCP server?

AI agents rely on the Model Context Protocol to discover and invoke server-side tools. A misconfigured endpoint — a missing capability declaration, an invalid JSON-RPC envelope, or an insecure CORS policy — can silently break tool discovery for every connected client. The MCP Inspector catches these issues before your users do, validating compliance against the MCP specification revision 2025-11-25 in under 15 seconds.

Who is it for?

Developers building MCP servers, platform teams integrating AI agents into production workflows, and security engineers auditing third-party tool providers. Whether you maintain a single-tool server or an enterprise gateway exposing hundreds of capabilities, the Inspector provides the same rigorous, specification-grounded analysis — completely free, with no authentication required.

What does the Inspector check?

Each scan executes 34 automated checks organized into 7 categories: DNS resolution and TLS certificate validation, JSON-RPC 2.0 protocol conformance, OAuth 2.1 authorization flow discovery, server security hardening (HTTPS enforcement, version leak detection, CORS policy), tool schema quality (input schema validation, naming conventions, unique tool names), trust and safety scanning (prompt injection, invisible Unicode, encoded payload detection), and performance benchmarking (initialize and tools/list latency thresholds).

How is the compliance grade calculated?

The overall score is a weighted average across all 7 categories, with each check mapped to a MUST, SHOULD, or MAY clause in the official MCP specification. Critical failures — such as plaintext HTTP or invalid JSON-RPC responses — immediately cap the maximum achievable grade. The final result is an A+ to F letter grade with a numeric score out of 100, accompanied by a per-check breakdown showing pass, fail, skip, or warning status with actionable remediation steps linked directly to the relevant specification section.

Three steps.
Under 15 seconds.

01

Paste your MCP server URL

Any Streamable HTTP endpoint. Public or authenticated. No install, no CLI, no sign-up.

02

34 automated compliance checks

Transport, protocol, OAuth 2.1, security, tool schemas, trust & safety, and performance.

03

Instant grade with remediation

A+ to F score. Every failed check includes a clear explanation and step-by-step fix.

34 checks.
7 categories.

Every check maps to a MUST, SHOULD, or MAY clause in the official Model Context Protocol specification.

MCP revision 2025-11-25. Transport, protocol, authorization, security, tool quality, trust and safety, and performance.

T Transport 4
DNS Resolution Validates hostname resolves to a routable IP address via A/AAAA lookup.
TLS Certificate Verifies a valid, non-expired TLS certificate is presented over HTTPS.
Endpoint Reachability Confirms the endpoint accepts connections and returns a well-formed HTTP response.
Response Content-Type Checks that the server responds with application/json as required by the MCP transport spec.
P Protocol 7
JSON-RPC 2.0 Format Validates the response envelope includes jsonrpc, id, and result or error fields.
Initialize Handshake Sends an initialize request and verifies the server returns capabilities and server info.
Protocol Version Confirms the negotiated protocolVersion matches a known MCP specification revision.
Session Management Checks for Mcp-Session-Id header presence when the server supports session tracking.
Capability Negotiation Verifies the server declares at least one capability (tools, resources, or prompts).
Ping/Pong Sends a ping request and validates the server responds within acceptable latency.
Error Format Triggers an invalid method call and verifies the error response follows JSON-RPC 2.0 structure.
A Authorization 5
Auth Discovery Sends an unauthenticated request and checks for a 401 response with WWW-Authenticate header per RFC 9728.
Protected Resource Metadata Fetches /.well-known/oauth-protected-resource and validates the required resource field.
Auth Server Metadata Retrieves /.well-known/oauth-authorization-server and verifies issuer, endpoints, and PKCE support.
PKCE S256 Support Confirms the authorization server advertises S256 in code_challenge_methods_supported per OAuth 2.1.
OAuth Endpoints Validates that authorization_endpoint, token_endpoint, and registration_endpoint are present and well-formed.
S Security 4
HTTPS Enforced Ensures the endpoint is served exclusively over TLS. Plaintext HTTP is a critical failure.
No Server Version Leak Checks response headers for framework or runtime version disclosure (Server, X-Powered-By).
CORS Policy Validates Access-Control-Allow-Origin is not set to wildcard (*) on authenticated endpoints.
No Error Info Leakage Inspects error responses for stack traces, file paths, or internal identifiers.
Q Tool Quality 6
Tool Discovery Calls tools/list and verifies the server returns a non-empty array of tool definitions.
Description Presence Checks every tool has a human-readable description field for AI model context.
Input Schema Defined Validates each tool declares an inputSchema for structured parameter validation.
Input Schema Type Object Ensures inputSchema.type is "object" as required by the MCP tool specification.
Unique Tool Names Detects duplicate tool names that could cause ambiguous tool selection by AI clients.
Naming Convention Verifies tool names follow lowercase snake_case or dot-separated notation.
X Trust & Safety 6
Prompt Injection Scan Scans tool descriptions for hidden instructions targeting the AI model (OWASP LLM01).
Invisible Unicode Detection Detects zero-width characters or bidirectional overrides used to conceal malicious directives.
Description Length Analysis Flags excessively long descriptions that may embed hidden payloads beyond visual inspection.
Encoded Payload Detection Searches for Base64, hex, or URL-encoded strings that could bypass content filters.
HTML Injection Detection Checks for HTML tags or script elements embedded in tool metadata fields.
Cross-Tool Reference Detects descriptions that instruct the model to invoke other tools, a tool-shadowing indicator.
F Performance 2
Initialize Latency Measures the time to complete the initialize handshake. Threshold: ≤ 5 000 ms.
Tools List Latency Measures tools/list response time. Threshold: ≤ 3 000 ms.