refactor: Update MCP tools to return human-readable text

[neptaco]

Summary

• Changed all MCP tool responses from structured JSON to formatted text output for improved readability in LLM interactions
• Updated grep tool to display search results with contextual log lines in a readable format
• Modified logs, ps, status, start, restart, and stop tools to use human-friendly text formatting
• All tools now consistently use MCP content blocks with type "text"

Changes by Tool

grep

• Replaced structured JSON matches with formatted text output
• Shows context before/after matched lines with clear visual indicators (<<< MATCH)
• Displays line numbers, timestamps, and content in a readable format

logs

• Changed from JSON array to formatted text with clear headers
• Added empty state handling with user-friendly message

ps

• Transformed process list from JSON objects to formatted text sections
• Separated processes with visual dividers
• Conditionally displays optional fields (PID, ports, start time)

status

• Converted detailed status information to readable text format
• Organized information hierarchically with clear labels
• Includes recent logs preview when available

start/restart

• Updated to show process information in formatted text
• Displays log context and matched lines in readable format
• Shows success indicators with checkmarks (✓)
• Improved timeout and failure messaging

stop

• Simplified output to show operation status and message

Test plan

• Test each MCP tool through the MCP interface
• Verify output formatting is consistent and readable
• Confirm all edge cases (no matches, no processes, failures) display appropriate messages
• Validate ANSI code stripping works correctly in log outputs

🤖 Generated with Claude Code