Endpoint report / Metadata + smoke test
DeepWiki MCP
A dated endpoint inspection. Metadata and any separately recorded tool call have their own evidence, scope and limitations.
What this establishes
The anonymous endpoint advertised 3 tools and passed four bounded metadata checks. A separate, fixed read-only smoke scenario is recorded below. Actual ChatGPT and Claude workflows were not tested.
The smoke result applies only to the displayed call and literal content assertions. It does not establish overall answer correctness or reliability.
- Connection and metadata
- Metadata checks passed
- Observed tool catalog
- Complete: 3 tools
- Read-only smoke test
- Passed · 1 call attempted
- ChatGPT / Claude
- Not tested / Not tested
- Overall score
- Not assigned
Read-only smoke test / Passed
Read the documentation structure for the public MCP TypeScript SDK repository.
This is one direct, fixed-input tool test. It does not exercise an AI model or a ChatGPT or Claude session. Calls attempted may include a call that returned an error or timed out.
The response passed the registered content assertions.
- Smoke tool calls attempted
- 1
- Tool
- read_wiki_structure
- Scenario / version
- deepwiki-sdk-structure / 1
- Observed at
- 2026-10-15T07:14:27.385Z
- Runner
- 1.0.0
- Protocol
- 2025-11-25
- Elapsed time
- 227 ms
- Run ID
- a5a04e58-5e73-4fc6-a76c-d0d026b14d56
- Result SHA-256
- e72d38e266509e4d6af6d167e83f3ec2de644b05d5455dfbfeba9e649bddadb3
Exact public inputs
{
"repoName": "modelcontextprotocol/typescript-sdk"
}Fixture assertions
- modelcontextprotocol/typescript-sdk: Passed
Recorded response excerpt
Available pages for modelcontextprotocol/typescript-sdk:
- 1 Overview
- 1.1 Installation and Setup
- 1.2 Monorepo Structure and Packages
- 1.3 Core Protocol Concepts
- 2 Architecture
- 2.1 Protocol Foundation
- 2.2 Type System and Message Schemas
- 2.3 Communication Lifecycle
- 2.4 Wire Codecs and Era-Aware Schemas
- 3 Server Implementation
- 3.1 McpServer High-Level API
- 3.2 Tool Registration and Execution
- 3.3 Resource and Prompt Management
- 3.4 Low-Level Server Protocol Class
- 3.5 Serving over HTTP
- 3.6 Middleware Integration
- 3.7 Serving over Stdio
- 3.8 Server Examples and Patterns
- 4 Client Implementation
- 4.1 Client Class and High-Level API
- 4.2 Streamable HTTP Client Transport
- 4.3 SSE Client Transport
- 4.4 Stdio and In-Memory Transports
- 4.5 Response Caching
- 4.6 Client Examples and Patterns
- 5 OAuth Authentication
- 5.1 OAuth Client Flow and Providers
- 5.2 Machine-to-Machine Authentication
- 5.3 OAuth Server Integration
- 6 Transport System
- 6.1 Transport Interface and Contract
- 6.2 Transport Comparison and Selection
- 7 Advanced Features
- 7.1 Session Management and Resumability
- 7.2 Elicitation a- One fixed public read-only request from a direct MCP client. No model selection, ChatGPT/Claude host, OAuth, private account or write behavior was tested.
- Content assertions check the listed literal terms and absence of a tool error; they do not establish complete factual correctness, relevance or production reliability.
- Observed call duration is one sample and includes network conditions. It is not a latency benchmark or a comparison between services.
- Only the first tools/list page is inspected for this selected tool. No notification stream, redirect, credentials or caller-supplied scenario is allowed.
- Response SHA-256 covers the UTF-8 JSON serialization of the SDK-parsed tool result. Only a bounded public text excerpt is retained; it is untrusted server content.
Recorded checks
These checks cover the metadata inspection. They are separate from the tool-call result above, when present.
protocol-negotiation
Negotiated protocol 2025-11-25; this does not establish host compatibility.
initialized-notification
The initialization notification received HTTP 202 with an empty body.
tool-metadata-shape
SDK metadata shape checked for 3 observed tools. This is not complete JSON Schema validation or behavior verification.
oauth-login
OAuth discovery, login, token refresh and authenticated access are outside this anonymous inspection.
Metadata request evidence
request-1 / initialize
2026-10-15T07:11:36.823Z
HTTP 200 · 1129 ms · 3829 B
response-received
request-2 / notifications/initialized
2026-10-15T07:11:37.952Z
HTTP 202 · 140 ms · 0 B
notification-accepted
request-3 / tools/list
2026-10-15T07:11:38.092Z
HTTP 200 · 183 ms · 1608 B
response-received
Observed tool catalog
Names were advertised through tools/list. Any tool execution is recorded separately in the smoke test section.
Complete: 3 tools
- ask_wiki_question
- read_wiki_contents
- read_wiki_structure
Endpoint and provenance
- Endpoint
- https://mcp.deepwiki.com/mcp
- Official source
- https://docs.devin.ai/work-with-devin/deepwiki-mcp
- Profile
- Anonymous endpoint profile
- Requested protocol
- 2025-11-25
- Reported protocol
- 2025-11-25
- Protocol
- 2025-11-25
- Observed at
- 2026-10-15T07:11:38.285Z
- Run ID
- b7e99b3f-5a4c-4edb-b0c6-5e92c74b8b72
- Catalog SHA-256
- 1cf852f73fb778b1c76a684ee85d786790a9e0705dfd2345f264602697432bd3
- Fingerprint scope
- complete-catalog
SHA-256 of all observed tool objects, recursively sorted object keys and sorted canonical tool strings; schema array order is preserved. Raw descriptions, schemas and headers are not persisted.
Inspection bounds
- requestTimeoutMs
- 5,000
- runTimeoutMs
- 30,000
- responseBytes
- 1,048,576
- requests
- 7
- pages
- 5
- tools
- 200
What this report does not establish
These observations do not establish general task success, security, OAuth behavior or ChatGPT and Claude compatibility. Read the limits of each recorded test.
- Metadata-only anonymous inspection. No tools/call, OAuth flow, API model calls, host UI or publication review was executed. No global quality or safety score is produced.
- Annotations are server declarations, not verified behavior. Metadata success does not establish task success, security, usefulness or ChatGPT/Claude compatibility.
- Only active exact HTTPS endpoints in the reviewed source registry are accepted. Redirects, supplied credentials, arbitrary endpoints and OAuth URL fetching are disabled. DNS/TLS use the local runtime: this is not a hardened multi-tenant SSRF sandbox and must not be exposed as a public scanner.
- Session identifiers are used only in memory. Request/response headers, raw error messages, raw descriptions and schemas are excluded from artifacts. Tool names and aggregate metadata remain untrusted server input.
- Only the recorded protocol versions and bounded catalog prefix are supported. Reaching a bound or lacking access means unknown, not a failed server or a passing full audit.
What should your MCP do better?
Share a public endpoint or repository, the tasks that matter, and the clients your users use. We will discuss a pilot scope, access and price.
Discuss a benchmarkOpens your email app. Send public links only; arrange sandbox access separately.