Skip to content
Back to search

ask-human-mcp

ask-human mcp πŸ§‘β€πŸ’»πŸ€πŸ€–

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/masony817--ask-human-mcp/badge.svg)](/listings/masony817--ask-human-mcp)

HTML

<a href="/listings/masony817--ask-human-mcp"><img src="/api/listings/masony817--ask-human-mcp/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

pip install ask-human-mcp
ask-human-mcp

`.cursor/mcp.json`:

{
  "mcpServers": {
    "ask-human": { "command": "ask-human-mcp" }
  }
}

restart cursor and vibe.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md Β· README Β· 226 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # ask-human mcp πŸ§‘β€πŸ’»πŸ€πŸ€–
  2. [![PyPI version](https://img.shields.io/pypi/v/ask-human-mcp?style=flat-square)](https://badge.fury.io/py/ask-human-mcp)
  3. [![License: MIT](https://img.shields.io/badge/License-MIT-yellow?style=flat-square)](https://opensource.org/licenses/MIT)
  4. [![Python 3.8+](https://img.shields.io/badge/python-3.8+-blue?style=flat-square)](https://www.python.org/downloads/)
  5. [!["Buy Me A Coffee"](https://img.shields.io/badge/buy_me_a_coffee-donate-orange?style=flat-square)](https://www.coff.ee/masonyarbrough)
  6. stop your ai from hallucinating. gives it an escape route when confused instead of false confidence.
  7. ## the pain
  8. ai blurts out an endpoint that never existed
  9. the agent makes assumptions that are simply not true and has false confidence
  10. repeat x100 errors and your day is spent debugging false confidence and issues when you could simply ask a question
  11. ## the fix
  12. an mcp server that lets the agent raise its hand instead of hallucinating. feels like mentoring a sharp intern who actually asks before guessing.
  13. agent β†’ ask_human()
  14. ⬇
  15. question lands in ask_human.md
  16. ⬇
  17. you swap "PENDING" for the answer
  18. ⬇
  19. agent keeps coding
  20. ### sample file:
  21. ```markdown
  22. ### Q8c4f1e2a
  23. ts: 2025-01-15 14:30
  24. q: which auth endpoint do we use?
  25. ctx: building login form in auth.js
  26. answer: PENDING
  27. ```
  28. you drop:
  29. ```markdown
  30. answer: POST /api/v2/auth/login
  31. ```
  32. boom. flow continues and hopefully the issues are solved.
  33. ## why it's good
  34. - pip install ask-human-mcp β†’ done
  35. - zero config, cross-platform
  36. - watches the file, instant feedback
  37. - multiple agents, no sweat
  38. - locks + limits so nothing catches fire
  39. - full q&a history in markdown (nice paper-trail for debugging)
  40. ## 30-sec setup
  41. ```bash
  42. pip install ask-human-mcp
  43. ask-human-mcp
  44. ```
  45. `.cursor/mcp.json`:
  46. ```json
  47. {
  48. "mcpServers": {
  49. "ask-human": { "command": "ask-human-mcp" }
  50. }
  51. }
  52. ```
  53. restart cursor and vibe.
  54. ## how it works
  55. 1. ai gets stuck β†’ calls `ask_human(question, context)`
  56. 2. question logged β†’ appears in `ask_human.md` with unique ID
  57. 3. human answers β†’ replace "PENDING" with your response
  58. 4. ai continues β†’ uses your answer to proceed
  59. the ai receives your answer and keeps coding!
  60. ## config options (if you want them)
  61. ### command line
  62. ```bash
  63. ask-human-mcp --help
  64. ask-human-mcp --port 3000 --host 0.0.0.0 # http mode
  65. ask-human-mcp --timeout 1800 # 30min timeout
  66. ask-human-mcp --file custom_qa.md # custom q&a file
  67. ask-human-mcp --max-pending 50 # max concurrent questions
  68. ask-human-mcp --max-question-length 5000 # max question size
  69. ask-human-mcp --rotation-size 10485760 # rotate file at 10mb
  70. ```
  71. ### different clients
  72. cursor (local):
  73. ```json
  74. {
  75. "mcpServers": {
  76. "ask-human": {
  77. "command": "ask-human-mcp",
  78. "args": ["--timeout", "900"]
  79. }
  80. }
  81. }
  82. ```
  83. cursor (http):
  84. ```json
  85. {
  86. "mcpServers": {
  87. "ask-human": {
  88. "url": "http://localhost:3000/sse"
  89. }
  90. }
  91. }
  92. ```
  93. claude desktop:
  94. ```json
  95. {
  96. "mcpServers": {
  97. "ask-human": {
  98. "command": "ask-human-mcp"
  99. }
  100. }
  101. }
  102. ```
  103. ## what's in the box
  104. - zero configuration β†’ works out of the box
  105. - file watching β†’ instant response when you save answers
  106. - timeout handling β†’ questions don't hang forever
  107. - concurrent questions β†’ handle multiple ai agents
  108. - persistent logging β†’ full q&a history in markdown
  109. - cross-platform β†’ windows, macos, linux
  110. - mcp standard β†’ works with any mcp client
  111. - input validation β†’ size limits and sanitization
  112. - file rotation β†’ automatic archiving of large files
  113. - resource limits β†’ prevent dos and memory leaks
  114. - robust parsing β†’ handles malformed markdown gracefully
  115. ## security stuff
  116. - input sanitization β†’ removes control characters and validates sizes
  117. - file locking β†’ prevents corruption from concurrent access
  118. - secure permissions β†’ files created with restricted access
  119. - resource limits β†’ prevents memory exhaustion and dos attacks
  120. - path validation β†’ ensures files are written to safe locations
  121. ## limits (so nothing breaks)
  122. | thing | default | what it does |
  123. |-------|---------|--------------|
  124. | question length | 10kb | max characters per question |
  125. | context length | 50kb | max characters per context |
  126. | pending questions | 100 | max concurrent questions |
  127. | file size | 100mb | max ask file size |
  128. | rotation size | 50mb | size at which files are archived |
  129. ## platform support
  130. - windows β†’ full support with native file locking
  131. - macos β†’ full support with fsevents file watching
  132. - linux β†’ full support with inotify file watching
  133. ## api stuff
  134. ### ask_human(question, context="")
  135. ask the human a question and wait for response.
  136. ```python
  137. answer = await ask_human(
  138. "what database should i use for this project?",
  139. "building a chat app with 1000+ concurrent users"
  140. )
  141. ```
  142. ### other tools
  143. - `list_pending_questions()` β†’ get questions waiting for answers
  144. - `get_qa_stats()` β†’ get stats about the q&a session
  145. ## development
  146. ### from source
  147. ```bash
  148. git clone https://github.com/masonyarbrough/ask-human-mcp.git
  149. cd ask-human-mcp
  150. pip install -e ".[dev]"
  151. ask-human-mcp
  152. ```
  153. ### tests
  154. ```bash
  155. pytest tests/ -v
  156. ```
  157. ### code quality
  158. ```bash
  159. black ask_human_mcp tests
  160. ruff check ask_human_mcp tests
  161. mypy ask_human_mcp
  162. ```
  163. ## contributing
  164. would love any contributors
  165. ### issues
  166. use the github issue tracker to report bugs or request features.
  167. you can also just email me: mason@kallro.com
  168. include:
  169. - python version
  170. - operating system
  171. - mcp client (cursor, claude desktop, etc.)
  172. - error messages or logs
  173. - steps to reproduce
  174. ## changelog
  175. see [CHANGELOG.md](CHANGELOG.md) for version history.
  176. ## license
  177. mit license - see [LICENSE](LICENSE) file for details.
  178. ## thanks
  179. - [model context protocol](https://github.com/modelcontextprotocol) for the excellent standard
  180. - [anthropic](https://anthropic.com) for claude and mcp support
  181. - [cursor](https://cursor.sh) for mcp integration
  182. - all contributors and users providing feedback

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.