The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →In Pennyforge Studio’s October 2026 probe, 40 of 78 registry-listed endpoints that answered initialization completed the full anonymous path of initialization, tool discovery and one safe read-style tool call: 51.3%. A separate study reported that 195 of 400 sampled npm/stdio packages completed a one-shot handshake: 48.8%. Both results are roughly half, but they measure different things in different samples. Neither is a reliability estimate for all MCP servers.
What “works” meant in the 78-endpoint probe
The headline refers to a bounded test reported by Pennyforge Studio on October 6, 2026. On October 2, the team examined 186 endpoints in one public registry’s alphabetical a–b slice. Of those, 87 returned HTTP 401; other endpoints errored or timed out before the test’s cohort was formed. The 78 endpoints that did answer an initialize call entered the reported cohort.
As an Amazon Associate I earn from qualifying purchases.
For those 78, the probe attempted an anonymous happy path: initialize, then tools/list, then one safe call. The tester selected the first declared tool with no required input properties and a read-style name, such as get, list or search. The reported outcomes were:
- 40 healthy: completed the full path, or 51.3% of the 78-endpoint cohort.
- 19 call errors: did not complete the selected tool call.
- 15 tool-list only: returned a tool list but did not complete the full path.
- 2 tool-list errors and 2 initialization errors.
That 51.3% is conditional on answering initialization and on the specific public registry slice. It is not the share of all 186 endpoints that worked, nor does it show whether endpoints that returned 401 would work with valid credentials. Pennyforge described 12-second caps, eight parallel workers and a test lasting about 20 minutes. The probe was a short snapshot, not a measurement of long-term uptime, production reliability, security or every tool each server exposes.
#1 Best Overall
- More for the money with this high quality Product
- Offers premium quality at outstanding saving
- Excellent product
- 100% satisfaction
Why another study also found a result near 50%
Masato Okuwaki’s October 2026 CORe Inc. article summarizes a September paper by Haseeb Mohammed Afsar, What a Random Draw from the MCP Registry Contains, and What Tool-Use Benchmarks Contain Instead (arXiv:2609.10962v1). The account describes a different sample and a narrower success test.
The study used a registry snapshot dated August 22, 2026. From 7,258 active npm packages declaring stdio transport within a reported registry population of 24,135 servers, it drew 400 packages without replacement, using a published seed and frame hash. Each package was launched once over stdio, without credentials or retries. The probe initialized, listed tools and validated schemas; it did not invoke a tool.
- 195 of 400 (48.8%) completed the handshake.
- 150 of 400 (37.5%) failed it.
- 53 of 400 (13.3%) required credentials.
- 2 of 400 (0.5%) were unavailable.
Because the probe allowed one attempt and no retry, a transient network or timing problem counted as a failure. Afsar therefore treats 48.8% as a floor for the share that might start, not as a point estimate of enduring availability. The tested frame was npm/stdio, representing 30.7% of the reported registry population; remote servers, PyPI packages and OCI images were not tested. Registry metadata is self-declared.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →These percentages answer different questions
| Probe | Sample and entry condition | Success condition | What the result does not establish |
|---|---|---|---|
| Pennyforge Studio, October 2026 | 78 endpoints from an a–b registry slice that had answered initialization | Anonymous initialization, tool listing and one selected safe read-style call; 40 of 78 (51.3%) completed | Authenticated success, long-term availability, or the reliability of all registry servers |
| Afsar study, as described by CORe Inc., October 2026 | 400 sampled active npm/stdio packages, launched once without credentials or retries | Initialization, tool listing and schema validation; 195 of 400 (48.8%) completed | Tool invocation success, retry behavior, or results for remote servers and other package ecosystems |
The Pennyforge percentage uses a cohort already selected for answering initialization and requires a tool call. The other percentage starts with a sample of npm/stdio candidates and counts a handshake without calling a tool. Their similar values do not make their numerators, denominators or success criteria interchangeable.
The CORe Inc. account also reports 66.7% handshake completion in a separate 24-server curated comparison frame under the same probe. That is a small curated frame, not evidence that curation itself causes better startup. It does illustrate why results from curated lists may differ from a broader random draw.
What an MCP health check should verify
MCP is an open protocol for connecting LLM applications with external data sources and tools. The Model Context Protocol specification (version 2025-11-25) describes JSON-RPC 2.0 messages between hosts, clients and servers, along with capability negotiation and server-provided tools, resources and prompts. A registry entry or a response to an initial request confirms only part of that interaction.
Rank #3
- Product type: Screw kit
- Made by Super Micro
- Manufacturer part number: MCP-410-00005-0N
- Supermicro MCP-410-00005-0N Screw Bag(100PCS) and Label for 24x Hot swap
- Mfr Part Number: MCP-410-00005-0N
For an integration team, a useful staged check is to verify the configured connection path before treating a server as ready for the application’s actual use:
Recommended Free Tools
- Launch or connect. Test the transport and deployment path the integration will use. A local npm/stdio launch is not a substitute for checking a remote endpoint.
- Initialize. Confirm that the server and client can complete protocol initialization and capability negotiation.
- List tools and inspect schemas. Check that discovery returns usable tool definitions and that input schemas can be validated. A successful tool list is not the same as a successful call.
- Exercise a tool only when safe and authorized. Use a read-only operation appropriate to the integration and the credentials available. Do not infer safety from a tool’s name alone.
- Repeat the check on a schedule and record outcomes. Afsar’s paper, as described by CORe Inc., recommends a scheduled handshake probe in CI. Repeated monitoring can reveal failures that a one-shot test cannot characterize, but the cited probes do not validate a particular monitoring strategy.
Anonymous discovery is not authenticated operation
A public, anonymous probe can identify what an unauthenticated client can reach, but it does not determine whether a credential-gated integration works after authorization. The 87 HTTP 401 responses in Pennyforge’s initial slice and the 53 credential-required packages in the separate sample are distinct outcomes, not proof that those servers are broken. A team should test with the credentials and permissions its intended deployment will actually use, while keeping public discovery results separate from authenticated checks.
Likewise, a one-shot check answers whether a server completed that attempt. It does not tell you how often it would recover on retry or whether it remains available over time. Retrying may be appropriate in a production monitor, but report first-attempt failures separately from eventual successes so transient and persistent problems are not conflated.
Startup failures are not security findings
The CORe Inc. account reports that 1,626 of 2,766 tools in the study’s corpus had no safety annotations. Missing annotations are missing information; they do not establish that a tool is either safe or dangerous. The paper explicitly makes no security claim, and neither handshake failures nor annotation gaps amount to a vulnerability rate.
A successful handshake is also not a security audit. It verifies protocol behavior within the probe’s scope, not the safety of a tool’s effects, access controls, data handling or implementation.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
