setup

Setup

Constraints

• MUST NOT use AskUserQuestion tool. Output plain text questions and wait for reply.
• MUST use Bash tool with python3 -c to write files (avoids Read/Write permission prompts).
• MUST write .mcp.json in the current project root (not ~/.mcp.json).

Steps

1. Print and wait for reply:

   ## Mono Memory Setup
   
   ---
   
   **Step 1/2** — Server URL
   
   Enter the full URL of your Mono Memory server:
   

2. Print and wait for reply:

   **Step 2/2** — Author
   
   Enter your name (used to tag memories you save):
   

3. Validate URL starts with http:// or https://. If not, ask to re-enter.

4. Detect project name from current directory name.

5. Run setup via Bash. MUST pass inputs as environment variables (not string interpolation):

   SERVER_URL='<URL>' AUTHOR_NAME='<author>' python3 -c "
   import json, os, sys
   
   url = os.environ['SERVER_URL']
   author = os.environ['AUTHOR_NAME']
   project = os.path.basename(os.getcwd())
   
   if not url.startswith(('http://', 'https://')):
       print('Error: URL must start with http:// or https://')
       sys.exit(1)
   
   # 1. Write .mcp.json
   mcp_path = os.path.join(os.getcwd(), '.mcp.json')
   config = {}
   if os.path.exists(mcp_path):
       with open(mcp_path) as f:
           config = json.load(f)
   config.setdefault('mcpServers', {})
   config['mcpServers']['mono-memory'] = {
       'type': 'http',
       'url': url
   }
   with open(mcp_path, 'w') as f:
       json.dump(config, f, indent=2)
   
   # 2. Append to CLAUDE.md
   claude_path = os.path.join(os.getcwd(), 'CLAUDE.md')
   template = '''
   
   ## Team Shared Memory (mono-memory MCP)
   
   ### Auto-Recording Rules
   When any of the following occur during work, **always** record them with \`memory_save\`:
   
   1. **Bug found & resolved**: root cause, fix, related files
   2. **Architecture/design decision**: why this approach, what alternatives were considered
   3. **New pattern/convention**: project-specific patterns discovered while coding
   4. **API changes**: endpoint additions/modifications, request/response format changes
   5. **DB/schema changes**: tables, columns, migrations
   6. **Dependency changes**: packages added/updated/removed and why
   7. **Environment/deployment changes**: env vars, build config, deployment pipeline
   8. **Performance issues**: bottleneck causes and improvements
   9. **Hard-won lessons**: problems that took significant time to solve (help others avoid the same)
   
   ### Recording Format
   When calling memory_save:
   - author: \"{author}\"
   - project: \"{project}\"
   - content: Always write in English, regardless of the conversation language. Describe specifically what was discovered/decided/resolved.
   - tags: relevant,comma,separated
   
   ### At Session Start
   Search for related records at the beginning of each session:
   - memory_search(query: \"keywords for your task\", project: \"{project}\")
   
   ### Check Project Context
   To review project structure or conventions:
   - memory_context(project: \"{project}\")
   
   ### Shortcut Commands
   When the user types \"/api-docs\", run the /mono-memory-mcp:api-docs skill.
   '''.format(author=author, project=project)
   
   existing = ''
   if os.path.exists(claude_path):
       with open(claude_path) as f:
           existing = f.read()
   
   if 'Team Shared Memory' in existing:
       print('CLAUDE.md already has memory config — skipped.')
   else:
       with open(claude_path, 'a') as f:
           f.write(template)
       print('CLAUDE.md updated.')
   
   print('Done!')
   "