Claude Code Integration

Introduction

Claude Code is Anthropic's official CLI tool that brings Claude directly to your terminal. With Lexicon5 integration, you can manage tasks, projects, and resources using natural language commands.

Unlike Windsurf which uses MCP configuration in settings.json, Claude Code uses a skill-based approach. You download a configuration package that includes skills, agents, and commands that integrate directly with your project's .claude folder.

Prerequisites

Before setting up the integration, ensure you have:

• Claude Code installed via npm
• A Lexicon5 account with an active project
• Node.js installed (for running task-manager.mjs)

npm install -g @anthropic-ai/claude-code

Setup

Follow these steps to configure Claude Code with Lexicon5:

# Extract to your project root
unzip lexicon5-claude-code.zip -d /path/to/your/project

# Verify the installation
ls -la .claude/

Configuration Package

The downloaded ZIP contains a complete configuration structure in the .claude folder:

.claude/
├── skills/
│   └── task-manager/
│       ├── .env              # Pre-configured credentials
│       ├── SKILL.md          # Skill documentation
│       ├── task-manager.mjs  # JavaScript script for API calls
│       └── references/
│           ├── Task-Template.md
│           ├── Example-Task-(RPG).md
│           └── tools.md
├── agents/
│   ├── lexicon5-developer.md
│   ├── lexicon5-qa-analyst.md
│   └── lexicon5-task-researcher.md
└── commands/
    ├── lexicon5-create-task.md
    ├── lexicon5-implement-task.md
    ├── lexicon5-issue-report.md
    └── lexicon5-init-nextjs-project.md

Environment Variables

The .claude/skills/task-manager/.env file contains your project credentials (pre-filled when downloaded):

Agents & Commands

The configuration package includes specialized agents and slash commands for workflow automation.

Agents

Agents are invoked automatically by Claude Code when relevant workflows are detected:

Slash Commands

Use these slash commands to trigger specific workflows:

Available Tools

The task-manager.mjs script provides direct access to Lexicon5 tools. Run commands from the .claude/skills/task-manager directory:

Task Operations

# List all tasks (default: only todo and in_progress)
node task-manager.mjs get_tasks

# Filter by status
node task-manager.mjs get_tasks '{"status":"backlog"}'

# Multiple statuses
node task-manager.mjs get_tasks '{"statusIn":["backlog","todo","in_progress","qa"]}'

# Get task with all details
node task-manager.mjs get_task '{"taskId":"task_abc"}'

# Get with subtask implementation details
node task-manager.mjs get_task '{"taskId":"task_abc","include":{"subtaskDetails":true}}'

# Create a new task
node task-manager.mjs create_task '{"title":"My task","description":"Task details"}'

# Update task status
node task-manager.mjs update_task '{"taskId":"task_abc","status":"in_progress"}'

# Add a comment to a task
node task-manager.mjs create_comment '{"taskId":"task_abc","content":"Progress update"}'

Subtask Operations

# Get subtask with implementation details
node task-manager.mjs get_subtask '{"subtaskId":"subtask_xyz"}'

# Mark subtask complete
node task-manager.mjs update_subtask '{"subtaskId":"subtask_xyz","completed":true}'

Tag Operations

# List all tags
node task-manager.mjs list_tags

# Create a new tag (find-or-create)
node task-manager.mjs create_tag '{"name":"Feature","color":"blue"}'

# Assign tag to task
node task-manager.mjs assign_tag '{"taskId":"task_abc","tagId":"tag_xyz"}'

Resource Operations

# List all resources
node task-manager.mjs --resources

# Read a specific resource
node task-manager.mjs --read-resource 'resource://project/xxx/resource/yyy'

# Create a resource
node task-manager.mjs create_resource '{"resourceTitle":"API Guidelines","content":"# Content..."}'

Token Optimization

Use the include parameter to control response size and reduce token usage:

Creating Tasks

You can create tasks in Lexicon5 using either natural language (free chat) or the dedicated slash command. Both approaches trigger the task-researcher agent to conduct discovery and create a comprehensive task.

Free Chat Approach

Simply describe what you need, and Claude will understand you want to create a task:

Slash Command Approach

Use the slash command to directly invoke the task creation workflow:

Implementing Tasks

The developer agent follows a structured workflow to implement tasks: load context, implement code, verify with lint/build, and hand off to QA.

Implementation Workflow

The developer agent follows these phases:

1. Load Context: Retrieves task details, subtasks, and implementation instructions
2. Implement: Works through each subtask, following project patterns
3. Verify: Runs pnpm lint and pnpm build to ensure quality
4. Hand Off: Adds completion comment with test plan, moves task to QA status

Reporting QA Issues

When you discover a bug, use the issue report command to create a structured bug task. The QA analyst agent researches your codebase to identify affected files and creates a comprehensive report.

What the QA Analyst Does

• Researches your codebase to understand the affected area
• Identifies specific files that may need modification
• Creates a structured bug task with reproduction steps
• Automatically applies the Bug tag
• Includes technical context for developers

Autonomous Project Manager

The autonomous project manager continuously monitors your task backlog and automatically spawns developer agents to work on tasks. This creates a hands-off development experience where work progresses without manual intervention.

How It Works

• Continuous monitoring: Scans for tasks in "todo" status
• Max concurrency: Works on 2 tasks simultaneously
• Progress checks: Reviews agent status every 5 minutes
• Idle mode: Waits 15 minutes when no tasks available
• Context management: Uses autocompact to manage context window

Example Log Output

The project manager provides detailed logging so you can monitor progress:

[LOOP START] Developer loop initialized at 2024-01-15 10:00:00
[SCAN] Checking for tasks in todo status...
[FOUND] 5 tasks in todo status
[SPAWN] Starting agent for task_abc123: "Add user settings page" (frontend)
[SPAWN] Starting agent for task_def456: "Fix API validation bug" (backend)
--- Waiting 5 minutes ---
[CHECK] Checking agent progress...
[COMPLETE] task_def456 completed successfully
[SCAN] Checking for tasks in todo status...
[FOUND] 4 tasks in todo status
[SPAWN] Starting agent for task_ghi789: "Update dashboard layout"
--- Waiting 5 minutes ---
[CHECK] Checking agent progress...
[COMPLETE] task_abc123 completed successfully
[COMPLETE] task_ghi789 completed successfully
[SCAN] Checking for tasks in todo status...
[FOUND] 0 tasks in todo status
[IDLE] No tasks in todo status. Waiting 15 minutes...

Slash Commands Reference

Complete reference of all available slash commands:

Free Chat Examples

You can interact with Lexicon5 using natural language. Here are examples of common queries and operations:

Querying Tasks

Filtering by Tags

Quick Status Updates

Project Overview

Free Chat vs Slash Commands

Understanding when to use each approach will help you work more efficiently with Lexicon5.

Comparison Table

Examples

Here are some realistic workflow examples using Claude Code with Lexicon5:

Creating a Feature Task

/lexicon5-create-task Add user profile page with avatar upload

Claude Code launches the task-researcher agent to conduct discovery, research the codebase, and create a comprehensive task with subtasks.

Implementing a Task

/lexicon5-implement-task task_abc123: Add user authentication

Claude Code launches the developer agent to retrieve task details, implement the feature, run verification, and move the task to QA.

Reporting a Bug

/lexicon5-issue-report Login form shows error on valid credentials

Claude Code launches the QA analyst agent to research the codebase, identify affected files, and create a structured bug report with proper tagging.

Conversation Examples

Task Lifecycle Workflow

A complete workflow from task creation to completion:

1. Create task: /lexicon5-create-task Add dark mode support
2. View task details: get_task with taskId
3. Start implementation: /lexicon5-implement-task task_xyz
4. Developer agent adds progress comments during development
5. Developer agent marks subtasks complete as work progresses
6. Developer agent moves task to QA status with test plan
7. QA verifies and marks task as done

Troubleshooting

Common issues and solutions: