Skip to content

FEATURES.md

Latest commit

 

History

History
873 lines (673 loc) · 21.3 KB

FEATURES.md

File metadata and controls

873 lines (673 loc) · 21.3 KB

CCG Features

Complete guide to all Code Guardian Studio features and 113+ MCP tools.

Overview

Code Guardian Studio (CCG) is the most comprehensive MCP server for Claude Code, offering:

  • 113+ MCP tools across 17 modules
  • Session persistence with automatic save/resume
  • Long-term memory across conversations
  • Real-time progress dashboard with WebSocket updates
  • Code analysis with hotspot detection
  • Workflow management with task tracking
  • Latent Chain reasoning for complex tasks

Pricing Tiers

CCG features are available across three tiers:

Tier Price Description
🆓 Dev Free Core features for individual developers
💎 Team $19/mo Advanced features for professional teams
🏢 Enterprise Custom SSO, audit logs, dedicated support

See Pricing for details.

Module Overview

Module Tools Tier Purpose
CCG Entrypoint 1 🆓 Natural language command interface
Session 10 🆓 Session lifecycle, export, resume
Memory 5 🆓 Persistent knowledge storage
Workflow 12 🆓 Task tracking and management
Guard 5 🆓 Code validation rules
Code Optimizer 8 🆓 Metrics, hotspots, refactoring
Process 8 🆓 Process and port control
Resource 10 🆓 Token budgeting, checkpoints
Progress 4 🆓 Real-time workflow visualization
Profile 8 🆓 Context profiles
Latent Chain 16 💎 Multi-phase reasoning
AutoAgent 20 💎 Task decomposition, error fixing
Thinking 10 💎 Reasoning models and workflows
RAG 6 💎 Semantic code search
Agents 7 💎 Multi-agent coordination
Testing 11 💎 Test runner, browser testing
Documents 9 💎 Documentation management

CCG Entrypoint

The /ccg command provides natural language access to all CCG tools.

ccg_run

Single entrypoint for natural language commands.

Usage:

/ccg "analyze code"
/ccg "run tests"
/ccg "check memory"

Parameters:

Parameter Type Description
prompt string Natural language command
dryRun boolean Preview without execution
translationMode enum auto, pattern, claude, tiny

Example Prompts:

  • "analyze code" - Run code analysis
  • "run tests" - Execute test suite
  • "check memory" - View memory summary
  • "validate code" - Run guard validation
  • "scan repo" - Full repository scan

Session Management

Manage sessions with automatic save and resume capability.

Tools

Tool Description
session_init Initialize new session, load memory
session_end End session, save all data
session_status Get current session info
session_timeline View session event history
session_export Export session to JSON
session_resume Resume from previous session
session_replay Replay timeline in dry-run mode
session_save Manual session save
session_offer Check for resumable sessions

Features

Automatic Session Management:

  • Sessions auto-save on significant events
  • Resume work after disconnects or crashes
  • Full context preservation including tasks and memory

Session Timeline:

  • Track all task completions
  • Record phase transitions
  • Create checkpoints at key moments

Export & Import:

// Export current session
session_export({ outputPath: "backup.json" })

// Resume from backup
session_resume({ sessionFile: "backup.json" })

Memory System

Persistent knowledge storage that survives across sessions.

Tools

Tool Description
memory_store Save information with tags and importance
memory_recall Search and retrieve memories
memory_list List all stored memories
memory_forget Remove specific memory
memory_summary Get memory statistics

Memory Types

Type Purpose Example
decision Choices made "Using React Query for data fetching"
fact Learned information "API uses JWT tokens"
code_pattern Reusable patterns "Error handling template"
error Mistakes to avoid "Don't use deprecated API"
note General notes "User prefers verbose logs"
convention Project rules "Use camelCase for variables"
architecture System design "Microservices with event bus"

Example Usage

// Store a decision
memory_store({
  content: "Using PostgreSQL for persistence",
  type: "decision",
  importance: 8,
  tags: ["database", "architecture"]
})

// Recall related memories
memory_recall({
  query: "database",
  type: "decision",
  minImportance: 5
})

Workflow Management

Track tasks, progress, and blockers throughout your development session.

Tools

Tool Description
workflow_task_create Create a new task
workflow_task_start Begin working on task
workflow_task_update Update progress
workflow_task_complete Mark task done
workflow_task_pause Pause current task
workflow_task_fail Mark task as failed
workflow_task_note Add notes to task
workflow_task_list List all tasks
workflow_task_delete Remove a task
workflow_current Get active task
workflow_status Workflow summary
workflow_cleanup Clean old tasks

Task States

pending -> in_progress -> completed
                      -> paused
                      -> blocked
                      -> failed

Example Workflow

// Create task
const task = workflow_task_create({
  name: "Implement auth",
  priority: "high",
  tags: ["feature", "security"]
})

// Start working
workflow_task_start({ taskId: task.id })

// Update progress
workflow_task_update({ taskId: task.id, progress: 50 })

// Add notes
workflow_task_note({
  taskId: task.id,
  content: "Using JWT with 1h expiry",
  type: "decision"
})

// Complete
workflow_task_complete({ taskId: task.id })

Guard Module

Validate code for quality issues, security problems, and bad patterns.

Tools

Tool Description
guard_validate Run validation rules on code
guard_check_test Check for fake tests
guard_rules List available rules
guard_toggle_rule Enable/disable rules
guard_status Module status

Available Rulesets

Ruleset Purpose
default General code validation
frontend React/Vue component rules
backend API/server code rules
security Security audit rules
testing Test file validation

Common Rules

Rule Detects
fake-tests Tests without assertions
empty-catch Empty catch blocks
disabled-features Commented/disabled code
inline-styles Inline CSS (frontend)
mixed-concerns Business logic in UI
sql-injection SQL injection vulnerabilities
xss Cross-site scripting risks

Example

guard_validate({
  code: fileContent,
  filename: "auth.ts",
  ruleset: "security",
  strict: true
})

Code Optimizer

Analyze codebases, find hotspots, and plan refactoring.

Tools

Tool Description
code_scan_repository Scan entire codebase
code_metrics Calculate file metrics
code_hotspots Find problem areas
code_refactor_plan Generate refactor steps
code_record_optimization Log optimization work
code_generate_report Create markdown report
code_quick_analysis All-in-one analysis
code_optimizer_status Module status

Metrics Calculated

Metric Description
Lines Total, code, comment, blank
Nesting Depth Maximum nesting level
Branch Score Complexity from conditionals
TODO/FIXME Pending work markers
Complexity Score Weighted composite

Hotspot Detection

Hotspots are files that need attention, ranked by:

  • Size: Large files that should be split
  • Complexity: High cyclomatic complexity
  • Nesting: Deep nesting levels
  • Mixed concerns: Files doing too much

Example Workflow

// Quick analysis (recommended)
const result = code_quick_analysis({
  strategy: "mixed",
  maxHotspots: 20
})

// Or step by step:
// 1. Scan repository
const scan = code_scan_repository()

// 2. Calculate metrics
const metrics = code_metrics({ files: scan.sourceFiles })

// 3. Find hotspots
const hotspots = code_hotspots({
  metrics,
  strategy: "complexity"
})

// 4. Generate plan
const plan = code_refactor_plan({
  hotspots,
  goal: "readability"
})

// 5. Generate report
code_generate_report({
  sessionId: "2024-01",
  scanResult: scan,
  hotspots,
  refactorPlan: plan
})

Testing Module 💎

Requires Team tier - Upgrade

Run tests and perform browser testing with Playwright.

Tools

Tool Description
testing_run Execute tests
testing_run_affected Run tests for changed files
testing_browser_open Open browser session
testing_browser_screenshot Take screenshot
testing_browser_logs Get console logs
testing_browser_network View network requests
testing_browser_errors Get JS errors
testing_browser_analysis Comprehensive analysis
testing_browser_close Close browser
testing_cleanup Clean test artifacts
testing_status Module status

Test Runner

// Run all tests
testing_run()

// Run with coverage
testing_run({ coverage: true })

// Run specific files
testing_run({ files: ["auth.test.ts"] })

// Run by pattern
testing_run({ grep: "login" })

Browser Testing

// Open browser
const session = testing_browser_open({
  url: "http://localhost:3000"
})

// Take screenshot
testing_browser_screenshot({
  sessionId: session.id,
  fullPage: true
})

// Get analysis
testing_browser_analysis({
  sessionId: session.id
})

// Close when done
testing_browser_close({ sessionId: session.id })

Latent Chain 💎

Requires Team tier - Upgrade

Multi-phase reasoning for complex tasks with context persistence.

Tools

Tool Description
latent_context_create Start new reasoning context
latent_context_get Retrieve context
latent_context_update Update with delta
latent_phase_transition Move to next phase
latent_apply_patch Apply code patch
latent_complete_task Finish task
latent_delete_context Remove context
latent_list_contexts List active contexts
latent_step_log Log reasoning step
latent_validate_response Validate response format
latent_diff_apply Apply unified diff
latent_diff_config Configure diff editor
latent_diff_confirm Confirm pending edit
latent_diff_pending List pending edits
latent_diff_rollback Rollback changes
latent_status Module status

Phases

analysis -> plan -> impl -> review
              ^       |
              +-------+  (go back if issues)

Context Structure

{
  taskId: "fix-auth-bug",
  phase: "analysis",
  codeMap: {
    files: ["src/auth.ts"],
    hotSpots: ["src/auth.ts:45"]
  },
  decisions: [
    { id: "D001", summary: "Use JWT", rationale: "Industry standard" }
  ],
  risks: ["Token expiry edge cases"],
  constraints: ["No breaking changes"]
}

Example Flow

// Create context
latent_context_create({
  taskId: "fix-auth",
  phase: "analysis"
})

// Log analysis step
latent_step_log({
  taskId: "fix-auth",
  phase: "analysis",
  description: "Found root cause in token validation"
})

// Transition to plan
latent_phase_transition({
  taskId: "fix-auth",
  toPhase: "plan",
  summary: "Analysis complete"
})

// Update with decisions
latent_context_update({
  taskId: "fix-auth",
  delta: {
    decisions: [{ id: "D001", summary: "Add token refresh" }]
  }
})

// Complete task
latent_complete_task({
  taskId: "fix-auth",
  summary: "Fixed token expiry bug"
})

AutoAgent 💎

Requires Team tier - Upgrade

Automatic task decomposition and error fixing.

Tools

Tool Description
auto_decompose_task Break task into subtasks
auto_analyze_complexity Score task complexity
auto_route_tools Suggest tools for action
auto_fix_loop Auto-fix errors
auto_fix_status Fix loop status
auto_store_error Save error pattern
auto_recall_errors Find similar errors
auto_create_graph Create task graph
auto_run_graph Execute task graph
auto_get_next_nodes Get next executable nodes
auto_start_node Start graph node
auto_complete_node Complete graph node
auto_fail_node Mark node failed
auto_graph_status Graph status
auto_list_graphs List all graphs
auto_delete_graph Delete graph
auto_workflow_start Start from template
auto_workflow_execute Execute workflow
auto_workflow_mermaid Visualize workflow
auto_agent_status Module status

Task Decomposition

auto_decompose_task({
  taskName: "Add authentication",
  taskDescription: "Implement JWT auth with login/logout",
  files: ["src/auth/"],
  constraints: ["Must be backward compatible"]
})

// Returns:
// - Complexity score (1-10)
// - Subtasks with dependencies
// - Suggested tools per subtask

Error Auto-Fix

auto_fix_loop({
  errorType: "TypeError",
  errorMessage: "Cannot read property 'map' of undefined",
  file: "src/components/List.tsx",
  maxRetries: 3
})

Agents Module 💎

Requires Team tier - Upgrade

Coordinate specialized agents for domain-specific tasks.

Tools

Tool Description
agents_list List available agents
agents_get Get agent details
agents_select Find best agent for task
agents_register Register new agent
agents_coordinate Coordinate multiple agents
agents_reload Reload agent definitions
agents_status Module status

Built-in Agents

Agent Specialization
trading-agent Financial/trading systems
laravel-agent Laravel/PHP development
orchestration-agent System orchestration

Agent Coordination

// Sequential: one after another
agents_coordinate({
  task: "Review and test auth",
  agentIds: ["code-reviewer", "test-runner"],
  mode: "sequential"
})

// Review: first does work, others review
agents_coordinate({
  task: "Implement feature",
  agentIds: ["developer", "reviewer", "tester"],
  mode: "review"
})

Documents Module 💎

Requires Team tier - Upgrade

Manage project documentation.

Tools

Tool Description
documents_search Search by query
documents_find_by_type Find by type
documents_should_update Check if update needed
documents_update Update existing doc
documents_create Create new doc
documents_register Register existing doc
documents_scan Scan for documents
documents_list List all docs
documents_status Module status

Document Types

  • readme - README files
  • spec - Specifications
  • api - API documentation
  • guide - User guides
  • changelog - Version history
  • architecture - System design
  • config - Configuration docs

Process Management

Manage processes and ports.

Tools

Tool Description
process_check_port Check if port available
process_check_all_ports Check configured ports
process_kill_on_port Kill process on port
process_kill Kill by PID
process_spawn Start new process
process_list List running processes
process_cleanup Kill all CCG processes
process_status Module status

Example

// Check if port is free
process_check_port({ port: 3000 })

// Kill process on port
process_kill_on_port({ port: 3000, force: true })

// Start dev server
process_spawn({
  command: "npm",
  args: ["run", "dev"],
  port: 3000,
  name: "dev-server"
})

Resource Management

Manage tokens, checkpoints, and resources.

Tools

Tool Description
resource_status Token usage stats
resource_update_tokens Update token count
resource_estimate_task Estimate task tokens
resource_checkpoint_create Save checkpoint
resource_checkpoint_list List checkpoints
resource_checkpoint_restore Restore checkpoint
resource_checkpoint_delete Delete checkpoint
resource_checkpoint_diff Compare checkpoints
resource_governor_state Get budget mode
resource_action_allowed Check if action allowed

Governor Modes

Mode Threshold Restrictions
normal < 70% None
conservative 70-84% Delta-only, no browser testing
critical >= 85% Checkpoint required, finish up

Progress Dashboard

Real-time workflow visualization.

Tools

Tool Description
progress_status Get workflow progress
progress_blockers List blocked nodes
progress_mermaid Generate Mermaid diagram
progress_clear Reset progress state

Mermaid Visualization

progress_mermaid({
  direction: "TD",  // top-down
  showGateBadges: true
})

// Returns Mermaid diagram code

RAG Module 💎

Requires Team tier - Upgrade

Semantic code search using embeddings.

Tools

Tool Description
rag_build_index Index codebase
rag_query Semantic search
rag_related_code Find similar code
rag_get_chunk Get code chunk details
rag_clear_index Clear index
rag_status Module status

Thinking Module 💎

Requires Team tier - Upgrade

Structured reasoning models and workflows.

Tools

Tool Description
thinking_get_model Get reasoning model
thinking_suggest_model Suggest model for task
thinking_list_models List all models
thinking_get_workflow Get SOP workflow
thinking_suggest_workflow Suggest workflow
thinking_list_workflows List workflows
thinking_save_snippet Save code template
thinking_get_style Get style reference
thinking_list_snippets List snippets
thinking_status Module status

Thinking Models

Model Use Case
chain-of-thought Step-by-step reasoning
tree-of-thoughts Multiple approaches
react Reasoning + Acting
self-consistency Verification
decomposition Break down problems
first-principles Fundamental analysis

Workflows

Workflow Purpose
pre-commit Before committing
code-review Reviewing code
refactoring Safe refactoring
deploy Deployment checklist
bug-fix Bug fixing process
feature-development New features
security-audit Security review

Profile Module

Manage context profiles for different projects or modes.

Tools

Tool Description
profile_create Create profile
profile_get Get profile details
profile_switch Switch active profile
profile_update Update profile
profile_delete Delete profile
profile_list List profiles
profile_detect Auto-detect profile
profile_status Module status

Quick Reference

Most Used Tools

Task Tool
Start session session_init
Analyze code code_quick_analysis
Run tests testing_run
Validate code guard_validate
Store knowledge memory_store
Create task workflow_task_create
Check tokens resource_status

Common Workflows

1. Code Analysis:

session_init -> code_quick_analysis -> code_generate_report

2. Bug Fix:

session_init -> latent_context_create -> latent_step_log ->
latent_phase_transition -> latent_complete_task

3. Test & Validate:

testing_run -> guard_validate -> workflow_task_complete

Related Documentation

Last updated: 2025-12-23