20 Best Code Execution MCP Servers, Compared (November 2026)

The short answer

Which one, for your situation

Top MCP servers for Code Execution

The ranking, with the evidence

• 1

  LLM Sandbox

  Community favourite

  vndee/llm-sandbox ↗

  Best for: Running LLM-generated code away from the host: it exposes three tools, including execute_code, and supports Docker, Kubernetes, Podman, Bubblewrap, NsJail and Windows sandbox backends.

  The server gives an MCP client three tools: execute_code, which runs code in a sandbox and captures visualizations, plus get_supported_languages and get_language_details for language discovery. Installation assumes a chosen isolation backend, so container SDK extras are needed for Docker, Kubernetes or Podman, and system executables for Bubblewrap and NsJail.

  Score 86.9: show every number behind it

  • Adoption100 / 100 · weight 40%
    • GitHub stars76
    • PyPI downloads100
    • Used through Glama14
  • Maintenance95 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence100
  • Momentum49 / 100 · weight 14%
    • Stars gained, relative to size43
    • Stars gained, absolute48
    • PyPI download trend62
  • Tool quality63 / 100 · weight 13%
    • Tool description quality55
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  86.9

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  86.9

• 2

  Jupyter MCP Server

  Community favourite

  datalayer/jupyter-mcp-server ↗

  Best for: Agents working in a live JupyterLab notebook: it exposes cell-level edit, execute and read tools, plus kernel restart and notebook activation.

  It connects to a Jupyter server by URL and token, then exposes 18 tools for listing files and kernels, activating a notebook, and inserting, editing, moving, clearing, executing and deleting cells, with a direct execute_code path for code that should not be saved to the notebook. It assumes a Jupyter instance already exists to attach to, and the README warns that code-sandboxes must be pinned to match the installed jupyter-mcp-server version, since a newer sandbox package with an older server fails on the first execution with an unknown sandbox variant error.

  Score 81.3: show every number behind it

  • Adoption78 / 100 · weight 40%
    • GitHub stars78
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum54 / 100 · weight 14%
    • Stars gained, relative to size53
    • Stars gained, absolute57
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality73 / 100 · weight 13%
    • Tool description quality65
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  81.3

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  81.3

• 3

  Context Mode

  Community favourite

  mksglu/context-mode ↗

  Best for: Coding agents that ingest large files or command output: it runs scripts in a sandbox and returns only what the script prints.

  Context Mode exposes 11 tools that run code in a sandboxed subprocess for JavaScript, shell, TypeScript, Python and Perl, index command output, files and fetched pages into a local SQLite knowledge base, and search that index. It installs as an npm package and keeps its indexes in a local SQLite database under ~/.context-mode, so sandboxed runs and search depend on locally installed language runtimes and local disk.

  Score 77.3: show every number behind it

  • Adoption100 / 100 · weight 40%
    • GitHub stars100
    • npm downloads93
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum100 / 100 · weight 14%
    • Stars gained, relative to size100
    • Stars gained, absolute100
    • npm download trend100
  • Tool quality93 / 100 · weight 13%
    • Tool description quality85
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  99.1

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  77.3

• 4

  Cloudflare MCP Server

  Community favourite

  cloudflare/mcp ↗

  Best for: Managing Cloudflare services by natural language chat without loading the API spec into context: it exposes search and execute tools that run code server-side.

  It connects over HTTP at mcp.cloudflare.com/mcp and exposes docs, search, execute, and whoami tools, where search and execute take JavaScript that queries the OpenAPI spec and calls cloudflare.request(). Before choosing it, note that in code mode each search and execute result is capped at roughly 6,000 tokens unless the call passes truncate: false, so results destined for a file or pipe need that flag.

  Score 65.8: show every number behind it

  • Adoption76 / 100 · weight 40%
    • GitHub stars76
    • npm downloadsnot measured
      npm names no repository for cloudflare-mcp, so its downloads cannot be attributed
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance95 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence100
  • Momentum95 / 100 · weight 14%
    • Stars gained, relative to size100
    • Stars gained, absolute86
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integrates100

  Weighted mean of the five

  84.4

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  65.8

• 5

  Containarium-Cloud

  Community favourite

  FootprintAI/Containarium ↗

  Best for: Agents that need a persistent Linux sandbox with SSH access and in-box command execution: the README documents an MCP server for shell_exec and file editing.

  Containarium is a self-hostable agent runtime that creates persistent sandboxes on LXC or Kubernetes and documents an in-box MCP server for shell_exec and file editing, plus CLI commands for creating boxes, syncing SSH configuration, and exposing ports. Self-hosting assumes a fresh Ubuntu 24.04 VM with sudo access for the installer, while a hosted cloud sign-up is also documented.

  Score 59.5: show every number behind it

  • Adoption63 / 100 · weight 40%
    • GitHub stars63
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum65 / 100 · weight 14%
    • Stars gained, relative to size76
    • Stars gained, absolute49
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  76.3

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  59.5

• 6

  Godot MCP - Full Control

  Community favourite

  tugcantopaloglu/godot-mcp ↗

  Best for: Driving a live Godot game from an AI assistant: game_eval runs GDScript inside the running game and returns values, alongside property, signal, and screenshot tools.

  It exposes 100 tools over MCP for the Godot engine, spanning project launch and debug output, headless scene and project settings editing, and runtime control such as game_eval, game_get_property, game_call_method, and game_screenshot against a running game. Before installing, note the requirements: Godot 4.4 or later, and distribution as the npm package @tugcantopaloglu/godot-mcp.

  Score 58.2: show every number behind it

  • Adoption71 / 100 · weight 40%
    • GitHub stars68
    • npm downloads21
      downloads show none of the weekday rhythm human traffic has; halved
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance83 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence40
  • Momentum75 / 100 · weight 14%
    • Stars gained, relative to size100
    • Stars gained, absolute67
    • npm download trend40
  • Tool quality53 / 100 · weight 13%
    • Tool description quality45
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  74.6

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  58.2

• 7

  lobu

  Steady

  lobu-ai/lobu ↗

  Best for: Agents that need durable shared context and governed sandboxed execution across company tools: it provides an event-sourced MCP context layer with persistent specialists.

  Lobu documents an MCP endpoint plus lobu connect CLI setup for Claude Code, Codex, OpenCode, Antigravity, ChatGPT, Claude Desktop, Cursor, and custom MCP clients, with OAuth and a shared event-sourced context layer for connected sources. Its MCP tool list is unknown, so exact operations cannot be confirmed before installation, and setup requires completing OAuth and connecting the sources whose context should be shared.

  Score 57.9: show every number behind it

  • Adoption59 / 100 · weight 40%
    • GitHub stars59
    • npm downloadsnot measured
      npm names no repository for lobu-monorepo, so its downloads cannot be attributed
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum60 / 100 · weight 14%
    • Stars gained, relative to size71
    • Stars gained, absolute44
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integrates100

  Weighted mean of the five

  74.2

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  57.9

• 8

  microsandbox

  Community favourite

  superradcompany/microsandbox ↗

  Best for: Running untrusted agent or user code in local microVMs: it provides hardware isolation, OCI image support, and fork/snapshot workflows.

  The README documents an msb CLI and TypeScript, Rust, Python, and Go SDKs for running commands in local microVMs from OCI images, with fork and snapshot support and a linked MCP server. It requires hardware virtualization enabled on the host (Apple Silicon on macOS, KVM on Linux, WHP on Windows) and is labeled beta software.

  Score 56.6: show every number behind it

  • Adoption99 / 100 · weight 40%
    • GitHub stars99
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum90 / 100 · weight 14%
    • Stars gained, relative to size84
    • Stars gained, absolute100
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  94.3

  × relevance: the keyword is tagged here

  0.60

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  56.6

• 9

  MCP Server Code Execution Mode

  Abandoned but popular

  elusznik/mcp-server-code-execution-mode ↗

  Best for: When an agent must call and compose existing stdio MCP server tools from Python in a rootless container: it exposes run_python and proxies those tools.

  The server exposes one run_python tool that executes Python in a persistent rootless container and proxies stdio MCP server tools so code can discover, call, and compose them. It assumes a Podman or Docker container runtime; no npm package is published and no PyPI package is attributed.

  Score 54.8: show every number behind it

  • Adoption63 / 100 · weight 40%
    • GitHub stars63
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance25 / 100 · weight 24%
    • Last commit touching this server42
    • Repository maintenance grade5
    • Commit cadence5
  • Momentum28 / 100 · weight 14%
    • Stars gained, relative to size31
    • Stars gained, absolute24
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality81 / 100 · weight 13%
    • Tool description quality73
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  54.8

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  54.8

• 10

  cmdshellmcp

  Community favourite

  ag88/cmdshellmcp ↗

  Best for: Constrained local automation in a disposable sandbox: it exposes 12 tools for allowlisted Unix commands, sandboxed file operations, patch application, and URL fetching.

  Its tools cover an allowlisted Unix command runner plus file read, write, list, copy, move, rename, and delete operations under a configured working directory, directory creation and removal, sed-based editing, patch application, and URL fetching. Before choosing it, note that it requires a configuration file and should run as an unprivileged user in a protected, disposable sandbox with authentication enabled by default.

  Score 53.8: show every number behind it

  • Adoption71 / 100 · weight 40%
    • GitHub stars12
    • npm downloadsnot measured
      no npm package
    • Used through Glama69
  • Maintenance95 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence100
  • Momentum0 / 100 · weight 14%
    • Stars gained, relative to size0
    • Stars gained, absolute0
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality68 / 100 · weight 13%
    • Tool description quality60
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  69.0

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  53.8

• 11

  deepseek-as-subagent

  Emerging

  PsChina/deepseek-as-subagent ↗

  Best for: Handing execution-heavy coding chores off from Claude Code or Codex CLI: the delegated model runs its own Read/Write/Edit/Bash loop inside a bounded workspace and returns a result.

  It exposes two tools, a ping health check and delegate_to_deepseek, the latter handing a focused task to DeepSeek or a local OpenAI-compatible model that then reads, edits, and runs commands in the configured workspace before returning its result to the host. Choosing it means supplying Python 3.10 through 3.12, a DeepSeek API key or a loopback OpenAI-compatible endpoint, and accepting that coding Bash runs on the trusted host with a workspace path boundary rather than in an operating-system sandbox.

  Score 49.1: show every number behind it

  • Adoption44 / 100 · weight 40%
    • GitHub stars44
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance100 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence100
  • Momentum81 / 100 · weight 14%
    • Stars gained, relative to size100
    • Stars gained, absolute52
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality96 / 100 · weight 13%
    • Tool description quality88
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  74.1

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: modest but real audience

  0.85

  Composite score

  49.1

• 12

  Damn Vulnerable Model Context Protocol (DVMCP)

  Community favourite

  harishsg993010/damn-vulnerable-MCP-server ↗

  Best for: Security researchers practicing MCP attack scenarios in a lab: it provides deliberately vulnerable challenge servers run via Docker, covering prompt injection and code execution.

  It packages deliberately vulnerable MCP challenge servers, started with Docker and exposed as Streamable HTTP endpoints on localhost ports. The documented setup relies on Docker, and the project warns that it is not stable in a Windows environment.

  Score 49.0: show every number behind it

  • Adoption78 / 100 · weight 40%
    • GitHub stars78
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance71 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade30
    • Commit cadence40
  • Momentum39 / 100 · weight 14%
    • Stars gained, relative to size36
    • Stars gained, absolute44
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust0 / 100 · weight 9%
    • License0
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  62.8

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  49.0

• 13

  Jupyter Server MCP

  Steady

  jupyter-ai-contrib/jupyter-server-mcp ↗

  Best for: Running a Jupyter Server and needing MCP clients to call configured Python functions as tools: it dynamically registers functions via module:function specifications.

  It is a Jupyter Server extension that runs a FastMCP streamable HTTP endpoint and registers Python functions as MCP tools via trait configuration or package entrypoints. It requires installation in the same environment as Jupyter Server or JupyterLab and depends on a running Jupyter Server for tool access.

  Score 43.6: show every number behind it

  • Adoption38 / 100 · weight 40%
    • GitHub stars38
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance97 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence85
  • Momentum40 / 100 · weight 14%
    • Stars gained, relative to size51
    • Stars gained, absolute24
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  62.3

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: small audience

  0.70

  Composite score

  43.6

• 14

  jupyter-editor-mcp

  Steady

  jsamuel1/jupyter-editor-mcp ↗

  Best for: Editing existing .ipynb files on disk without running a Jupyter server: it exposes 29 tools for reading, modifying and batch-processing notebook cells.

  The server exposes 29 tools that read, search, insert, replace, delete, reorder, filter and validate notebook cells, and that merge, split, search across, sync metadata for and clear outputs from multiple .ipynb files, each tool taking file paths rather than a live kernel connection. It edits notebook files only: no tool executes code or runs cells, so actually running a notebook still requires a separate Jupyter installation.

  Score 40.9: show every number behind it

  • Adoption37 / 100 · weight 40%
    • GitHub stars19
    • PyPI downloads34
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance88 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence65
  • Momentum29 / 100 · weight 14%
    • Stars gained, relative to size38
    • Stars gained, absolute15
    • PyPI download trendnot measured
      no download history for the selected registry
  • Tool quality73 / 100 · weight 13%
    • Tool description quality65
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  58.4

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: small audience

  0.70

  Composite score

  40.9

• 15

  EdgeBox

  Abandoned but popular

  BIGPPWONG/EdgeBox ↗

  Best for: Agents that need to operate an Ubuntu desktop: it documents VNC GUI automation with screenshots, mouse, and keyboard control plus shell and code tools.

  EdgeBox is a local sandbox that documents MCP interfaces for shell commands, multi-language code interpretation, filesystem access, and VNC desktop automation. It assumes a local Docker container and desktop application setup, and at publication no npm package was published and no PyPI package was attributed.

  Score 40.9: show every number behind it

  • Adoption58 / 100 · weight 40%
    • GitHub stars58
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance46 / 100 · weight 24%
    • Last commit touching this server68
    • Repository maintenance grade30
    • Commit cadence5
  • Momentum0 / 100 · weight 14%
    • Stars gained, relative to size0
    • Stars gained, absolute0
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  52.4

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  40.9

• 16

  mcp-sandbox-computer-vm-for-ai

  Steady

  flujo-app/mcp-sandbox-computer-vm-for-ai ↗

  Best for: Agents that need isolated shell execution on named Linux computers: terminal_execute runs bash in a Debian sandbox with a stable computer_id for reattachment.

  The server exposes one tool, terminal_execute, which runs bash commands in an isolated Debian Linux sandbox and returns stdout, stderr, exit code, duration, computer_id, and temporary status. The lifecycle tools and dashboard are opt-in: set ENABLE_LIFECYCLE_TOOLS=true before starting the server if named-computer management beyond terminal_execute is needed.

  Score 40.9: show every number behind it

  • Adoption33 / 100 · weight 40%
    • GitHub stars15
    • PyPI downloads30
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance97 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade100
    • Commit cadence85
  • Momentum0 / 100 · weight 14%
    • Stars gained, relative to size0
    • Stars gained, absolute0
    • PyPI download trendnot measured
      no download history for the selected registry
  • Tool quality100 / 100 · weight 13%
    • Tool description quality93
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  58.4

  × relevance: the keyword is dedicated here

  1.00

  × continuity: actively changing

  1.00

  × evidence: small audience

  0.70

  Composite score

  40.9

• 17

  open-skills

  Abandoned but popular

  instavm/open-skills ↗

  Best for: Processing sensitive files locally on an Apple Silicon Mac: it runs Claude Skills in a sandbox using Apple's native containers and exposes an MCP endpoint.

  The README documents a local MCP server at http://open-skills.local:8222/mcp, with setup examples for Claude Desktop, OpenCode, Gemini CLI, and OpenAI Agents, while the exposed tool list has not been inspected. Before choosing it, note that it requires an Apple Silicon Mac with macOS and Python 3.10+, and installation is via cloning the repository and running install.sh.

  Score 40.8: show every number behind it

  • Adoption66 / 100 · weight 40%
    • GitHub stars66
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance32 / 100 · weight 24%
    • Last commit touching this server42
    • Repository maintenance grade30
    • Commit cadence5
  • Momentum0 / 100 · weight 14%
    • Stars gained, relative to size0
    • Stars gained, absolute0
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  52.3

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: widely adopted

  1.00

  Composite score

  40.8

• 18

  Mobile Device MCP

  Dormant

  srmorete/mobile-device-mcp ↗

  Best for: Controlling multiple iOS and Android devices simultaneously from an AI agent: it discovers devices and proxies tap, screenshot, UI tree, and run_code calls per device.

  Mobile Device MCP documents tools to list devices, capture screenshots, read UI trees, tap, scroll, type text, press hardware buttons, launch and terminate apps, install apps, and run sandboxed JavaScript on iOS and Android, with calls proxied to on-device servers. One dependency to note is Node.js plus platform tooling: adb for Android, xcrun/simctl for iOS Simulator, and xcodebuild, devicectl, and iproxy for iOS real devices.

  Score 40.4: show every number behind it

  • Adoption47 / 100 · weight 40%
    • GitHub stars43
    • npm downloads22
      downloads show none of the weekday rhythm human traffic has; halved
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance69 / 100 · weight 24%
    • Last commit touching this server88
    • Repository maintenance grade80
    • Commit cadence5
  • Momentum54 / 100 · weight 14%
    • Stars gained, relative to size78
    • Stars gained, absolute37
    • npm download trend33
  • Tool quality≈70 / 100 · weight 13%
    • Tool description quality≈70
      tool descriptions not yet scored
    • Built and inspected by Glamanot measured
      never built and inspected by Glama
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  60.9

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: modest but real audience

  0.85

  Composite score

  40.4

• 19

  MCP Python Interpreter

  Steady

  yzfly/mcp-python-interpreter ↗

  Best for: Local Python scripting in a designated working directory: it lists environments and packages, runs code or files, installs packages, and reads and writes files.

  MCP Python Interpreter exposes 10 tools for listing Python environments and installed packages, executing inline code or Python files, installing packages, reading and writing files, listing Python files in a directory, and clearing or listing REPL sessions. One caveat is that file tools and run_python_code(save_as=...) restrict paths to the configured working directory unless MCP_ALLOW_SYSTEM_ACCESS=1, and inline or subprocess execution runs arbitrary Python with the server user's permissions and is not an operating-system sandbox.

  Score 39.9: show every number behind it

  • Adoption56 / 100 · weight 40%
    • GitHub stars51
    • PyPI downloads35
      PyPI downloads show no weekday rhythm; halved
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance76 / 100 · weight 24%
    • Last commit touching this server100
    • Repository maintenance grade80
    • Commit cadence5
  • Momentum10 / 100 · weight 14%
    • Stars gained, relative to size0
    • Stars gained, absolute0
    • PyPI download trend42
  • Tool quality71 / 100 · weight 13%
    • Tool description quality63
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  60.2

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: modest but real audience

  0.85

  Composite score

  39.9

• 20

  Civil3D MCP Server

  Dormant

  barbosaihan/civil3d-mcp ↗

  Best for: Automating Autodesk Civil 3D drawings from an AI assistant: it exposes C# code execution, read-only query, and documented skill-template browsing.

  Civil3D MCP Server provides three MCP tools: civil3d_execute runs C# code with write access in a committed transaction, civil3d_query runs C# code in read-only mode, and civil3d_skills browses documented C# code templates. Before choosing it, note that it requires a running Civil 3D plugin reachable by the MCP server and build steps for both the MCP server and the plugin.

  Score 39.2: show every number behind it

  • Adoption42 / 100 · weight 40%
    • GitHub stars42
    • npm downloadsnot measured
      no npm package
    • Used through Glamanot measured
      not used through Glama in the last 30 days
  • Maintenance46 / 100 · weight 24%
    • Last commit touching this server68
      dated from the last commit on the default branch, re-read from GitHub at publication; github.com shows a push 79 days ago, which counts every ref; the stored date would have published 79 days ago
    • Repository maintenance grade30
    • Commit cadence5
  • Momentum73 / 100 · weight 14%
    • Stars gained, relative to size94
    • Stars gained, absolute42
    • npm download trendnot measured
      no download history for the selected registry
  • Tool quality93 / 100 · weight 13%
    • Tool description quality85
    • Built and inspected by Glama100
  • Trust100 / 100 · weight 9%
    • License100
    • Published by the vendor it integratesnot measured
      not published by the vendor it integrates

  Weighted mean of the five

  59.1

  × relevance: the keyword is declared here

  0.78

  × continuity: actively changing

  1.00

  × evidence: modest but real audience

  0.85

  Composite score

  39.2

Questions people ask

Other comparisons

• Best Figma MCP servers
• Best Obsidian MCP servers
• Best YouTube MCP servers
• Best Jira MCP servers
• Best PostgreSQL MCP servers
• Best MySQL MCP servers
• Best GitHub MCP servers
• Best Notion MCP servers
• Best Linear MCP servers
• Best Discord MCP servers
• Best MongoDB MCP servers
• Best CRM MCP servers
• Best Payments MCP servers
• Best Ecommerce MCP servers
• Best Social Media MCP servers
• Best Browser Automation MCP servers
• Best Email MCP servers
• Best PDF MCP servers
• Best Kubernetes MCP servers
• Best Docker MCP servers
• Best AWS MCP servers
• Best Spreadsheets MCP servers
• Best SEO MCP servers
• Best Web Search MCP servers
• Best Web Scraping MCP servers
• Best Vector Database MCP servers
• Best Calendar MCP servers
• Best Observability MCP servers
• Best Image Generation MCP servers
• Best Speech & Transcription MCP servers
• Best Maps MCP servers
• Best Travel MCP servers
• Best Stock Market MCP servers
• Best Web Analytics MCP servers
• Best Memory MCP servers
• Best Documentation MCP servers
• Best Security MCP servers