Top 10 MCP Servers for Other Utilities in 2026

The "Other" category of the MCP Trust Registry is where the genuinely essential utilities live — the servers that do not fit a neat vertical like databases or messaging, but that end up in almost every serious agent config anyway. Structured reasoning, persistent memory, live documentation, web fetching, timezone math, and controlled access to your local machine: these are the building blocks that turn a chat model into an agent that can plan, remember, and act.

Because these servers are so widely installed, they are also a large attack surface. A memory server that reads and writes files, or a desktop tool that runs shell commands, deserves more scrutiny than its star count alone provides. That is the differentiator for every entry below: each one has been statically scanned by MCP Trust Checker, so you see its A–F Trust Score — a deterministic 0–100 rating of the actual published source — alongside what it does.

How to read the grade in one sentence: an A (90–100) means our engine found little to worry about, while lower grades or a non-zero count of high-severity findings mean there are specific behaviours — filesystem writes, command execution, network calls — worth reviewing on the server's registry page before you grant it access. Here are the ten we recommend, popularity-led and best-first.

1. Context7

Context7 pulls version-specific, up-to-date documentation and code examples for thousands of libraries and injects them straight into the model's context. Instead of generating code against a stale training snapshot, the model sees the API of the version you actually have installed — you just add use context7 to a prompt. With roughly 20k GitHub stars, it is consistently the top-recommended docs-and-context server in current roundups.

MCP Trust Score: A (100/100) · GitHub · npm

2. Sequential Thinking

An official reference server from the modelcontextprotocol project, Sequential Thinking gives a model a structured scratchpad for step-by-step reasoning. It can lay out a chain of thoughts, revise earlier steps, branch into alternatives, and adjust how many steps it needs mid-task. For agents that would otherwise answer complex problems in a single shot, it is one of the simplest ways to improve multi-step planning — which is why it appears on nearly every "awesome MCP" list.

MCP Trust Score: A (100/100) · GitHub · npm

3. Desktop Commander

Desktop Commander turns a chat client into a lightweight coding-and-automation agent. It gives the model controlled access to your machine: running terminal commands with streaming output and long-running process management, plus surgical file reads, edits, and search across the filesystem. With millions of npm downloads it is one of the most popular community servers, and a fixture of "best MCP servers" lists.

Be deliberate with this one. Its power is exactly what our scanner flags: it is a high-capability server, and our scan reports 11 high-severity findings tied to its command-execution and filesystem access. That does not make it malicious — it is doing what it advertises — but review those findings on its registry page and understand the blast radius before you let an agent run it unattended.

MCP Trust Score: B (83/100) · GitHub · npm

4. Memory (Knowledge Graph)

The official reference implementation of persistent memory for MCP. Memory stores entities, relations, and observations in a knowledge graph that survives across sessions, so an assistant can remember facts about you and your projects instead of starting cold every conversation. It is the canonical, heavily-forked baseline that most other memory servers are measured against.

MCP Trust Score: A (100/100) · GitHub · npm

5. Fetch

Fetch is the minimal, dependency-light way to give a model read access to the open web. It retrieves a URL, converts the HTML to clean Markdown, and supports chunked reading so long articles fit in context — no headless browser or scraping stack required. As an official reference server it is the default "let the model read a web page" tool and one of the most-installed MCP packages on PyPI.

MCP Trust Score: A (98/100) · GitHub · PyPI

6. Time

Small but genuinely essential: Time provides current-time lookups and timezone conversions using IANA timezone names, with automatic detection of the system timezone. It fixes a real blind spot — models have no reliable internal clock — which makes it valuable for scheduling, "what time is it in X", and any relative-time reasoning. It shows up in almost every starter MCP config.

MCP Trust Score: A (98/100) · GitHub · PyPI

7. Blender MCP

Blender MCP connects Blender to an LLM through a Blender addon plus an MCP server, enabling prompt-assisted 3D modeling. You can create and manipulate objects, apply materials, inspect the scene, and run Python inside Blender — describing a scene in natural language and watching it built in the viewport. At around 24.7k GitHub stars it is the breakout creative server and a favourite showcase of what MCP can do beyond text.

One honest note: our scan flags 1 high-severity finding, which is unsurprising given that the server can execute arbitrary Python inside Blender. Review it on the registry page and treat scene scripts as you would any code you run locally.

MCP Trust Score: C (76/100) · GitHub · PyPI

8. Basic Memory

Basic Memory is a local-first alternative to the reference Memory server. It stores notes as plain Markdown files in a Zettelkasten-meets-knowledge-graph style, and gives assistants bidirectional, semantic access to them. Because the data is human-readable files on your own disk, you and the AI edit the same notes and nothing is locked in a proprietary store — which has made it popular with the Obsidian and personal-knowledge-management crowd.

Worth reading carefully: despite an overall A, this is a high-capability server and our scan reports 10 high-severity findings connected to its read/write access to your notes and filesystem. The grade reflects that the code is clean and well-structured; the findings reflect what it is allowed to do. Check them on the registry page before pointing it at a sensitive vault.

MCP Trust Score: A (92/100) · GitHub · PyPI

9. Everything

Everything is the official protocol demonstration server, and it earns its place on this list for developers rather than end users. It deliberately exercises every part of the MCP spec — tools, prompts, resources, sampling, and notifications — which makes it the standard sandbox for building or debugging MCP clients and servers. If you are writing an MCP integration, this is the conformance target you run to verify full-protocol support.

MCP Trust Score: A (100/100) · GitHub · npm

10. Airbnb

A practical, self-contained example of an MCP server for real-world tasks: it searches Airbnb listings with filters and returns detailed property information, letting an assistant help plan trips and compare stays. It works without an official API key by using public endpoints, which makes it easy to try — and a commonly cited demonstration of MCP handling everyday, non-developer tasks.

MCP Trust Score: A (98/100) · GitHub · npm

How we graded these

Every score above comes from the deterministic, open-source MCP Trust Checker engine. It reads the actual source published to npm or PyPI — not a README, not a marketing page — and applies a fixed set of rules to model each server's capabilities and toxic-flow risks. There is no LLM in the scoring path, so the same input always produces the same score: the grades here are reproducible, not opinions. A high-severity finding is not an accusation of malice; it flags a capability (running commands, writing files, making network calls) that increases what could go wrong, so you can make an informed decision. Browse the full methodology and per-server reports at the registry.

This is a curated slice of a larger category — browse every scanned server in the full "Other" category, and scan any npm or PyPI package yourself for free via the MCP Trust Checker API before you add it to an agent.

npx mcptrustchecker