--- title: Advanced Features description: Complex features and multi-agent coordination --- ### What You'll Learn This section covers Verdent's power-user capabilities including subagent management, rule systems, custom tool integration, and advanced extensibility options. --- ## Complex Feature Explanations Verdent for VS Code provides sophisticated features that enable advanced development workflows beyond basic code generation. ### Subagent System **Specialized AI Agents:** Verdent includes built-in subagents (`@Verifier`, `@Explorer`, `@Code-reviewer`) that operate with isolated context windows and custom system prompts. Each subagent specializes in specific tasks like code exploration, validation, or quality review. **Automatic & Manual Delegation:** The main agent automatically routes tasks to appropriate subagents based on request patterns, or users can manually specify subagents using @-mentions. This delegation preserves main context while handling specialized operations efficiently. **Custom Subagents:** Create project-specific subagents by defining markdown files with YAML frontmatter in `~/.verdent/subagents/`. Custom subagents can specialize in domain expertise, team workflows, or technology-specific optimizations. [Learn more about subagent management](/docs/verdent-for-vscode/agents-rules/subagent-management) ### Rule Systems **Behavior Customization:** Rule files guide Verdent's behavior, output formatting, and decision-making through plain Markdown documents. Three rule types control different scopes: - **VERDENT.md** - Global personal preferences across all projects - **AGENTS.md** - Project-specific team standards (version controlled) - **plan_rules.md** - Plan Mode output format customization **Rule Precedence:** When conflicts occur, project rules (`AGENTS.md`) override user rules, ensuring team consistency while preserving personal preferences for solo projects. [Learn more about rule systems](/docs/verdent-for-vscode/agents-rules/rule-systems) ### Tool System & MCP **Built-in Tools:** Verdent provides comprehensive tools for file operations, search (`glob`/`grep`), command execution (`bash`), web access, and agent delegation. Tools enable all code manipulation, navigation, and external interaction. **Model Context Protocol (MCP):** MCP extends Verdent's capabilities by connecting to external tools, data sources, and services. Configure MCP servers through Settings → MCP Servers to integrate project-specific APIs and custom functionality. [View tool reference](/docs/verdent-for-vscode/advanced-features/tool-reference) [Learn about MCP integration](/docs/verdent-for-vscode/advanced-features/mcp) --- ## Multi-Agent Coordination Verdent's multi-subagent architecture enables parallel task execution with automatic routing and coordination. **Main Agent (Orchestrator):** The primary agent analyzes requests, breaks down complex tasks, and delegates specialized work to subagents while maintaining conversation context and consolidating results. **Parallel Execution:** Multiple subagents run concurrently for complex operations. Example: Explorer subagent searches the codebase while Code-reviewer analyzes security simultaneously, delivering faster results. **Context Isolation:** Each subagent operates with its own context window. Only final results consume space in the main agent's context, preventing exhaustion during multi-step operations. **Task Routing:** Automatic detection routes tasks based on patterns: - Codebase questions → `@Explorer` subagent - Security reviews → `@Code-reviewer` subagent - Validation checks → `@Verifier` subagent --- ## Custom Workflow Creation Advanced users can customize Verdent's behavior to match team processes and project requirements. **Custom Subagent Workflows:** Create specialized subagents for domain-specific needs: - **API documentation generators** - Extract endpoints and parameters from code - **Database migration reviewers** - Validate schema changes for safety - **Accessibility auditors** - Check frontend code for WCAG compliance **Rule-Based Customization:** Define coding standards, testing requirements, and output preferences through rule files: - Enforce team conventions (indentation, naming, patterns) - Specify documentation styles and comment detail levels - Control Plan Mode output structure and granularity **Extensibility Through MCP:** Integrate external tools and services: - Database query tools - Cloud service APIs (AWS, Azure, GCP) - Project management integrations (Jira, Linear) - CI/CD pipeline connections [Learn about custom subagents](/docs/verdent-for-vscode/agents-rules/subagent-management) [Explore extensibility options](/docs/verdent-for-vscode/advanced-features/extensibility) --- ## Advanced Context Management Techniques Efficient context management is critical for complex projects and long coding sessions. **Strategic File Reading:** Use `file_read` with line ranges for large files to avoid context overload. Read only sections needed for the task rather than entire files. **Subagent Delegation for Research:** Delegate exploratory tasks to `@Explorer` subagent to keep main conversation context clean. Subagent research results consume minimal main context space. **Search Before Reading:** Use `grep_file` to identify relevant files before reading full contents. Combine `glob` patterns with `grep_content` for precise searches that minimize unnecessary file operations. **Context-Aware Tool Selection:** - **Large codebases:** `@Explorer` subagent for architecture understanding - **Isolated validation:** `@Verifier` subagent for quick checks - **Security reviews:** `@Code-reviewer` subagent for quality assessment **Monitoring Context Usage:** Track which subagents are executing through Chat View indicators. Subagent operations appear in conversation history with progress indicators and result attribution. --- ## See Also Complete tool capabilities guide Custom subagents, rules, and MCP integration