The MCP server ecosystem
Five zones covering the Model Context Protocol surface — the reference servers Anthropic maintains, the vendor-owned servers, the community work worth knowing, the hosts that consume them, and the spec itself. Read /systems/mcp first if you need the protocol-engineering depth before scanning the landscape.
Anthropic reference servers
The set Anthropic maintains in modelcontextprotocol/servers. These are the canonical implementations — what every third-party MCP server gets compared to for argument validation, transport handling, and capability negotiation. Start here before reaching for a community port.
MCP Filesystem Server
Anthropic's reference MCP server for filesystem access. Read, write, search, move, and list files inside a configured allowlist of directories. The canonical example for understanding how MCP tool exp
MCP Fetch Server
Reference MCP server for fetching and converting web content. Pulls a URL, runs HTML through a readability extractor, returns markdown the model can chunk and reason over. The lightweight web-reader p
MCP Git Server
Reference MCP server for local Git repository operations. Status, diff, log, blame, branch listing — read-side operations against a checked-out repo without round-tripping to GitHub. Pairs with mcp-se
MCP Memory Server
Reference MCP server that gives an agent a persistent knowledge graph — entities, relations, observations stored to disk and surfaced back across sessions. The simplest path to making an agent remembe
MCP Brave Search Server
Reference MCP server wrapping the Brave Search API. Privacy-respecting alternative to Google/Bing endpoints — Brave does not maintain a personal-history-linked index. The default web-search MCP in the
MCP Sequential Thinking
Reference MCP server that gives an agent a structured scratchpad for multi-step reasoning. Each call records a numbered thought with revision and branching support — the agent can backtrack, fork, and
MCP PostgreSQL Server
Reference MCP server that exposes a Postgres database as a query surface. Read-only by default — but worth flagging that early versions had a SQL-injection class issue where the read-only wrapper coul
Vendor-maintained servers
Servers maintained by the platform they wrap. GitHub took over its own MCP server in late 2025; Microsoft owns Playwright MCP. The signal here is durability — vendor-owned servers track the underlying API on the vendor's release cadence, not the community's.
MCP GitHub Server
GitHub's first-party MCP server. Surfaces issues, pull requests, code search, file contents, repo metadata, Actions runs, and discussions through the protocol. Now maintained by GitHub itself rather t
Playwright MCP
Microsoft's MCP server that drives a real browser via Playwright — Chromium, Firefox, and WebKit. Ships ~22 tools that operate against the page's accessibility tree rather than pixel coordinates, whic
Community standouts
The independent servers that survive the registry's long tail. We list the ones we've actually verified — not just the registry top-of-page. Selection bar: clear ownership, real test surface, an actual operational reason to use them over the reference set.
Clients (hosts)
The other side of the protocol. Without a host that speaks MCP, none of the servers above do anything. Coverage in the host space is uneven — Claude Desktop and Claude Code are the strictest implementations; agent runtimes like OpenHands and Goose treat MCP as one of several tool transports.
Claude Desktop
Anthropic's official desktop app for Claude. Native MCP server support means you can plug in local file access, GitHub, and custom tools. Distinct from the Claude Code CLI.
Claude Code
Anthropic's terminal-native coding agent. Tops SWE-bench Verified at 87.6% and SWE-bench Pro at 64.3% in 2026. Deep MCP integration, agentic file editing, and a $20/mo Pro tier are the standout signal
Cursor
Anysphere's AI-native IDE. Forks VS Code with Cursor Tab inline completion, agentic chat, and background agents. Best 'flow' for inline completion in 2026.
OpenHands
AI-driven development agent that completes engineering tasks end-to-end — branches, code, PRs. v1.6 added a Planning Mode that drafts a plan before executing. Local-LLM-friendly via Ollama, vLLM, and
Goose
Open-source extensible AI agent now governed by the Agentic AI Foundation (AAIF) at the Linux Foundation. Started inside Block (formerly Square). 25+ provider support including Ollama, Ramalama, Docke
Cline
VS Code extension agent — ~4M installs in 2026. Plan/Act mode, autonomous file edits with diff approval, terminal access. The leading open-source IDE agent.
Continue
Open-source VS Code and JetBrains assistant. Configurable autocomplete + chat + agent modes. Strong with local Ollama backends.
Protocol layer
The spec itself. MCP is a JSON-RPC 2.0 dialect with a specific lifecycle, capability negotiation, and a schemaful tool/resource/prompt surface. Read the system guide before deploying anything in production.
Remote MCP — what shipped, what to watch
The protocol's first eighteen months were dominated by local stdio servers — single-user, single-machine, configured per project. That centre of gravity is moving. Between March and May 2026, Atlassian, HubSpot, Linear, Slack, Sentry, Neon, and Vercel all shipped first-party remote MCP endpoints; the production count went from ~16 to 25+ in eight weeks. Three things follow from that. First, the host-side question changes: your client has to handle OAuth and rotate session tokens, not just fork a subprocess. Second, the threat model changes: the prompt-injection surface is no longer "what's on this laptop" — it's whatever payload the remote service decides to return. Third, registries start mattering. We track the public-facing surface here, but the enterprise-deployed long tail is rapidly outgrowing what any one map can show. We'll add a registry-snapshot subpage when the numbers force it.
For protocol-level depth — capability negotiation, the local stdio versus remote HTTP+SSE split, the security model, the latency budget — see /systems/mcp. That guide is the upstream of this map.
How this map updates
This page reads its zones live from the catalog. When a new MCP server lands in scripts/seed/tools.ts and references one of the zone slugs above, it shows up here on the next deploy. Editorial framing — zone titles, blurbs, the "what changed this month" callout, the remote-MCP narrative — is hand-written and refreshed on the first business day of each month. We don't include every server that lands in the public registry; the bar is "we've actually verified ownership and surface" rather than "it scored on Glama this week."
Going deeper
- What MCP is really solving — the protocol guide. Lifecycle, tool invocation flow, security model, latency budget, local vs remote tradeoffs.
- The local AI agent ecosystem — the wider map this one slots into. MCP is one zone of the agent landscape; that page shows how it relates to coding agents, runners, and memory frameworks.
- MCP protocol catalog entry — the spec entity itself with version history, transports, and client/server matrix.