meta-cc
Meta-cognition tool for Claude Code - Analyze session history, detect patterns, optimize workflows.
Note: Skills and agents from previous versions have been moved to yaleh/baime. meta-cc 3.0.0 focuses exclusively on session history analysis via MCP tools.
What is meta-cc?
meta-cc helps you understand and improve your Claude Code workflows through:
- Autonomous analysis - Claude automatically queries session data via MCP tools
- 21 MCP tools - Error analysis, quality scanning, work patterns, timelines, bug detection, and more
- Prompt library - Save, search, and reuse optimized prompts with 3 slash commands
Zero configuration required - works out of the box with Claude Code.
Quick Install
Method 1: Plugin Marketplace (Recommended)
/plugin marketplace add yaleh/meta-cc
/plugin install meta-cc
Restart Claude Code — that's it. The MCP server is automatically configured via .mcp.json bundled in the plugin.
The meta-cc plugin includes:
- 3 Slash Commands -
/prompt-find, /prompt-list, /prompt-show for prompt library management
- 21 MCP Tools - Session data analysis with two-stage query architecture (v2.1)
Method 2: Archive Install (Alternative)
# Linux/macOS (one-liner)
curl -L https://github.com/yaleh/meta-cc/releases/latest/download/meta-cc-plugin-linux-amd64.tar.gz | tar xz
cd meta-cc-plugin-linux-amd64
./install.sh
The archive installer copies the binary and integration files, and automatically merges the MCP server configuration into ~/.claude/mcp.json.
Other platforms: See Installation Guide for macOS (Apple Silicon), Windows, and manual installation.
Verify Installation
In Claude Code, ask naturally:
"Show me all Bash errors in this project"
"Which tools do I use most often?"
"Find user messages mentioning 'refactor'"
Troubleshooting: See Installation Guide for common issues.
Quick Start
Autonomous Analysis (MCP)
Just ask Claude naturally - MCP tools are invoked automatically:
"Show me all Bash errors in this project"
"Find user messages mentioning 'refactor'"
"Which tools do I use most often?"
"Scan session quality and show me scores"
"Show my work patterns and peak hours"
"Find bug fix pairs in my session"
Unified query interface with 21 MCP tools and jq filtering:
// Core query tool - unified interface
query({
resource: "tools",
filter: {tool_status: "error"},
jq_filter: '.[] | select(.tool_name == "Bash")'
})
// Convenience tools - optimized for common queries
query_tool_errors({limit: 10})
query_token_usage({stats_first: true})
query_conversation_flow({scope: "session"})
// Raw jq - maximum flexibility for power users
query_raw({
jq_expression: '.[] | group_by(.tool_name) | map({tool: .[0].tool_name, count: length})'
})
// New analysis tools (3.0.0)
analyze_errors({}) // Aggregate errors by tool and type
quality_scan({}) // Compute error/retry/diversity scores
get_work_patterns({}) // Hourly activity and context switches
get_timeline({}) // Chronological session events
analyze_bugs({}) // Error-fix pairs and recurring patterns
get_tech_debt({}) // TODO/FIXME markers and unresolved errors
Key Features:
- Hybrid Output Mode: Auto-switches between inline (<8KB) and file_ref (≥8KB)
- jq Integration: Native jq filtering for complex queries
- No Limits by Default: Returns all results, relies on hybrid mode
- 21 Tools: 2 core + 8 convenience + 7 legacy + 3 utility + 6 analysis tools
Resources:
Prompt Library (Slash Commands)
Save and reuse your best prompts with 3 built-in commands:
/prompt-find phase execution # Search by keywords
/prompt-list sort=usage # Browse all (sorted by use)
/prompt-show phase-execution-001 # View full prompt details
Documentation
Getting Started