MCP servers
MCP servers, reviewed for what they can reach.
Model Context Protocol servers give an AI agent tools: files, a browser, a database, GitHub, live docs. Each profile here states the security scope, the install channel, the credentials, a config entry you can paste, and how far we have verified it ourselves.
Evidence graded — hands-on verified, source validated, or cataloged
Security scope up front
Read-only, read-write or local execution — stated on every profile before the install block.
Upstream state, not memory
Deprecated, archived and unpublished packages are detected from a committed upstream snapshot and labeled.
Graded evidence
Hands-on Verified means we ran it; Source Validated means we read the source and docs. 5 of 29 profiles are hands-on verified today.
Start here
Before you connect a server
Three short reads that make the profiles below easier to judge.
Learn
What is MCP, in plain English?
How a client, a server and a model fit together, and what a server can reach once it is connected.
Read the guide →Secure
MCP security best practices
Scope credentials, keep approval prompts on, and know which servers can execute code on your machine.
Read the checklist →Build
MCP agent workflow stack
A tested combination of a local runtime, an MCP-capable client and a small set of servers.
Open the stack →Editorial review
Grades describe our evidence, not the server's quality. A Source Validated profile has not been run by us; verify the install and credential scope yourself before granting an agent access.
Showing 29 of 29 MCP servers.
Documentation + context
AWS Documentation MCP
Search and read official AWS documentation through MCP without giving the agent credentials to your AWS account.
View profile →Infrastructure
AWS MCP Server
Search AWS knowledge and run IAM-authorized AWS API workflows through AWS's managed preview MCP server.
View profile →Memory + knowledge
Basic Memory MCP
Give agents durable local memory in plain Markdown, with search, graph traversal and explicit note creation, editing and deletion.
View profile →Browser + DevTools
Chrome DevTools MCP
The Chrome DevTools team's MCP server: DOM snapshots, console, network and performance traces from a live Chrome as structured tool results.
View profile →Data + databases
ClickHouse MCP
Query ClickHouse databases through the official MCP server, with read-only SQL enforced by default and separate write opt-ins.
View profile →Infrastructure
Cloudflare API MCP
Search and operate the Cloudflare API through a compact Code Mode MCP server with OAuth-scoped account access.
View profile →Documentation + context
Cloudflare Docs MCP
Give MCP agents current Cloudflare documentation context without granting any access to your Cloudflare account.
View profile →Memory + knowledge
Codebase Memory MCP
A native, local MCP server that indexes a repository into a persistent knowledge graph — functions, classes, call chains, routes — for fast queries.
View profile →Documentation + context
Context7
Upstash's MCP server that puts current, version-specific library docs and code examples in the prompt so a coding agent stops guessing at APIs.
View profile →Documentation + context
DeepWiki MCP
Read generated documentation and ask source-grounded questions about public GitHub repositories through DeepWiki.
View profile →Data + databases
DuckDB / MotherDuck MCP
Query local DuckDB files or MotherDuck through MCP, with read-only defaults and optional write-enabled SQL.
View profile →Web + search
Exa MCP
Search the web and fetch clean page content through Exa, with optional advanced search and multi-step research tools.
View profile →Web + search
Fetch
Reference MCP server that fetches a web page and returns it as Markdown, so a local model can read live content without a browser.
View profile →File system
Filesystem
Reference MCP server for local file operations — read, search, edit and move files inside the directories you explicitly allow.
View profile →Web + search
Firecrawl MCP
Search, scrape, parse and crawl the live web through Firecrawl, with optional browser interaction and autonomous research.
View profile →Developer tools
Git
Reference MCP server for reading, searching and changing a local Git repository — status, diffs, log, commits, branches and checkouts.
View profile →Developer tools
GitHub MCP Server
GitHub's official MCP server: repositories, issues, pull requests, code search and Actions as tools — hosted remotely or run locally in Docker.
View profile →Data + databases
Hugging Face MCP
Search Hugging Face models, datasets, Spaces, papers and docs through MCP, with optional Hub and compute tools.
View profile →Memory + knowledge
Memory
Reference MCP server that gives an agent persistent memory as a local knowledge graph — entities, relations and observations stored in a JSONL file.
View profile →Infrastructure
Netlify MCP
Build, deploy and manage Netlify sites and account resources through Netlify's official hosted MCP server.
View profile →Memory + knowledge
Notion MCP
Search, read and update authorized Notion workspace content through Notion's official hosted MCP server.
View profile →Browser + DevTools
Playwright MCP
Microsoft's MCP server that drives a real browser through Playwright with accessibility snapshots, not screenshots — navigate, click, type, read.
View profile →Memory + knowledge
Qdrant MCP
Give agents a local or hosted Qdrant memory layer for storing notes and retrieving them by semantic search.
View profile →Data + databases
Redis MCP
Let agents read, write, search and manage Redis data structures through Redis's official MCP server.
View profile →Developer tools
Semgrep MCP
Run Semgrep security scans through MCP so coding agents can analyze source code for vulnerabilities locally.
View profile →Developer tools
Sentry MCP
Inspect errors, traces and issues, then perform scoped issue triage through Sentry's official MCP service.
View profile →Utility
Sequential Thinking
Reference MCP server that gives a model a structured, revisable step-by-step reasoning tool — no files, no network, nothing to secure.
View profile →Data + databases
Supabase MCP
Supabase's official MCP server: lets an agent manage tables, run SQL, read logs and deploy Edge Functions in your project — scope it read-only.
View profile →Web + search
Tavily MCP
Search, extract, map and crawl live web content through Tavily's official hosted MCP server.
View profile →How to read the grades
- Hands-on Verified Installed, discovered and executed in a recorded environment; the profile lists each configuration.
- Source Validated Install channel, config, credentials and license checked against the source and upstream metadata; not run by us yet.
- Cataloged Known to exist; no profile until it is validated. Listed below for completeness.
Cataloged, not yet profiled (9)
These servers are known to the directory but have no profile yet — several are deprecated or archived upstream. They stay listed so you can see what we have looked at, without a page that could go stale.
- Brave Search (archived reference server)CatalogedArchived upstreamThe MCP reference Brave Search server. Its npm package is deprecated; the reference repository says it has been replaced by Brave's official server.
- Docker (community)CatalogedA community Docker-management MCP server (Python, run with uvx). Cataloged only: an unrelated npm package shares its name.
- Everything (test server)CatalogedThe MCP reference test server that exercises every protocol feature — for people building MCP clients, not for real work.
- GitLab (archived reference server)CatalogedArchived upstreamThe MCP reference GitLab server. Its npm package is deprecated and the reference repository lists it under Archived.
- MySQL (community)CatalogedA community MySQL/MariaDB MCP server. Cataloged only: the npm package this directory pointed at states no repository, so provenance is unresolved.
- PostgreSQL (archived reference server)CatalogedArchived upstreamThe MCP reference PostgreSQL server (read-only schema inspection and queries). npm package deprecated; listed under Archived upstream.
- Puppeteer (archived reference server)CatalogedArchived upstreamThe MCP reference Puppeteer browser server. npm marks its package deprecated and the reference repository lists it under Archived.
- Slack (archived reference server)CatalogedArchived upstreamThe MCP reference Slack server. npm package deprecated; listed under Archived upstream, which names Zencoder’s server as the continuation.
- SQLite (archived reference server)CatalogedArchived upstreamThe MCP reference SQLite server (a Python package run with uvx). The reference repository lists it under Archived; its last release was April 2025.
Next step
Pick a client, then a server.
A server is only useful inside a client that speaks MCP. Start with the client you already use, then add one server at a time and keep approval prompts on.