servers vs ha-mcp
Two of the top projects in MCP servers, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
servers
Reference MCP servers for files, git, fetch, memory and time
ha-mcp
MCP server that lets AI assistants configure and control Home Assistant
Verdict
servers ranks higher, 94 to 80. Of the 4 score parts we measure for both, servers leads on 1, ha-mcp on 1 and 2 are even.
- Adoptionservers leads by 57 points, 98 to 41: a wide gap.
- FreshnessEven at 100.
- MaintenanceEven at 99.
- Agent-readyha-mcp leads by 40 points, 85 to 45: a wide gap.
Pick servers if you are
Developers learning MCP or wiring basic tools into an MCP client
- Maintained by the MCP steering group, so it tracks the spec and official SDKs
- Seven small servers covering files, git, web fetch, knowledge-graph memory and time
Pick ha-mcp if you are
Home Assistant users who want MCP clients to manage their smart home
- 87 tools covering automations, dashboards, helpers, areas, backups, traces and device registry
- In-process integration works on OS, Supervised, Container and Core installs
| What we compare | servers | ha-mcp |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 98, widely used | 41, known |
| Momentum | Not scored for this kind | Not scored for this kind |
| Freshness | 100, active | 100, active |
| Maintenance | 99, healthy | 99, healthy |
| Agent-ready | 45, minimal | 85, ready |
| Facts from GitHub and the README | ||
| Stars | 91.9k | 7.1k |
| License | custom license (read the license) | MIT (permissive) |
| Last commit | Nov 2026 | Nov 2026 |
| Last release | Nov 2026 | Nov 2026 |
| Language | TypeScript | Python |
| Docker | No | Yes |
| GPU | Not needed | Not needed |
| arm64 or Apple Silicon | Not stated | Mentioned |
servers
Official collection of reference Model Context Protocol servers maintained by the MCP steering group: Everything, Fetch, Filesystem, Git, Memory, Sequential Thinking and Time. TypeScript servers start with npx and Python ones with uvx or pip. All serve stdio; only Everything also serves Streamable HTTP. The README states they are educational examples, not production-ready.
Who it is for: Developers learning MCP or wiring basic tools into an MCP client
Strengths
- Maintained by the MCP steering group, so it tracks the spec and official SDKs
- Seven small servers covering files, git, web fetch, knowledge-graph memory and time
- Starts with one command: npx for TypeScript servers, uvx or pip for Python ones
- Everything server exercises prompts, resources and tools, and also serves Streamable HTTP
Weaknesses
- README says these are reference implementations, not production-ready
- Only Everything supports Streamable HTTP; the other servers are stdio only
- Many earlier servers (GitHub, Postgres, Slack, Redis and others) are archived
- License is reported as NOASSERTION: Apache-2.0 for new contributions, MIT for older code
- no GPU
- Needs Node.js (npx), uv or pip (Python servers), MCP client
ha-mcp
ha-mcp exposes 87 tools over the Model Context Protocol so clients such as Claude, ChatGPT and Cursor can search entities, control devices, and create or edit automations, scripts, helpers and dashboards in Home Assistant. It is built on FastMCP and installs as an in-process HACS/Marketplace integration, a Home Assistant app (add-on), Docker, or uvx/PyPI. Safety features include read-only mode, per-tool disabling and automatic edit backups.
Who it is for: Home Assistant users who want MCP clients to manage their smart home
Strengths
- 87 tools covering automations, dashboards, helpers, areas, backups, traces and device registry
- In-process integration works on OS, Supervised, Container and Core installs
- Read Only Mode, per-tool enable/disable and automatic backups before edits
- Setup Wizard generates config for 15+ clients; demo server available to try first
Weaknesses
- Integration requires Home Assistant 2026.8.0 or newer
- Acts with an administrator token; a leaked token or secret URL grants broad access
- stdio transport has known connection problems; README recommends it only for demos
- Several install paths exist; running more than one per client causes connection hangs
- no GPU
- Docker + Compose
- Needs Home Assistant 2026.8.0+, Home Assistant administrator access token, FastMCP
- Models: Claude, ChatGPT, Gemini CLI, Open WebUI, Cursor
- port 9584
More in MCP servers
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.