Worker System

Production-ready automated task execution using AI agents.

Overview

The AnyTask Worker System enables automated task execution using AI agents. Workers run continuously, monitoring for tasks that match workflow triggers, then automatically implementing them using Claude Code. Most users should use the built-in workflows - they’re production-ready and cover common development scenarios. For custom automation needs, see the Complete Guide for writing your own workflows.

Key Features

Built-in Workflows - Four production-ready workflows included with CLI
AI-Powered Execution - Uses Claude Code for intelligent task implementation
Event-Driven - Triggers on task events (created, updated, status changed)
Zero Configuration - Works out of the box with sensible defaults
Error Handling - Graceful failure handling with automatic notifications

Built-in Workflows

The CLI includes four production-ready workflows. Simply start the worker and they’re available immediately - no configuration needed.

1. Dry Run (`dry_run`)

Purpose: Test workflow execution without making real changes When to use: Testing worker setup, validating workflows, CI/CD validation Triggers:

Any task created or updated to todo status

What it does:

Simulates task analysis with Claude
Simulates implementation (no actual file changes)
Simulates linting and testing
Simulates git commit
Marks task as done
Posts detailed simulation logs to task

Timeout: 10 minutes Benefits:

Fast execution
No API costs (uses dry-run actions)
Safe for testing
Validates workflow orchestration

# Use dry run for testing
anyt worker start --agent-id agent-xxx --workflow dry_run

2. Local Development (`local_dev`)

Purpose: Quick implementation on current repository state When to use: Rapid iterations, local testing, simple fixes Triggers:

Any task created or updated to todo status

What it does:

Analyzes task with Claude Haiku
Posts analysis to task
Implements changes with Claude Code
Runs linting (make lint)
Runs tests (make test)
Commits all changes
Marks task as done

Timeout: 30 minutes Model: claude-haiku-4-5-20251001 Key Features:

No branch switching (works on current state)
Fast execution with Haiku model
Continues even if lint/test fail
Perfect for iterative development

# Best for: Quick iterations on current branch
anyt worker start --agent-id agent-xxx --workflow local_dev

3. Feature Development (`feature_dev`)

Purpose: Full feature implementation with PR creation When to use: New features that need code review Triggers:

Tasks with feature label, status todo

What it does:

Creates feature branch (feature/TASK-123-...)
Analyzes task requirements
Posts implementation plan
Implements feature with Claude Code
Commits changes
Pushes branch to remote
Creates GitHub pull request
Updates task with PR link
Marks task as done

Timeout: 60 minutes Model: claude-haiku-4-5-20251001 Key Features:

Automatic branch creation
GitHub PR creation with detailed description
PR includes task identifier for tracking
Dependency caching for faster execution

# Best for: Feature development with PR workflow
anyt worker start --agent-id agent-xxx --workflow feature_dev

# Then create a task with feature label:
anyt task add "Add dark mode toggle" --labels feature --status todo

4. Feature with Quality Control (`feature_dev_with_check`)

Purpose: Enterprise-grade feature development with quality gates When to use: Production features requiring quality assurance Triggers:

Tasks with feature label, status todo

What it does: Pre-flight Checks:

Verifies make check command exists (or lint/typecheck/test)
Creates blocking task if quality commands missing
Checks master branch is clean
Creates cleanup task if uncommitted changes found
Runs quality checks on master
Creates fix task if quality checks fail

Implementation (only if all checks pass):

Creates feature branch
Analyzes task with quality considerations
Implements feature
Runs post-implementation quality checks
Fails if quality checks don’t pass
Commits only if all checks pass
Marks task as done

Timeout: 90 minutes Model: claude-haiku-4-5-20251001 Key Features:

Automatic quality gate setup
Blocks on quality issues
Creates dependency tasks automatically
Ensures clean master branch
Post-implementation validation
Enterprise-ready quality control

# Best for: Production features with quality requirements
anyt worker start --agent-id agent-xxx --workflow feature_dev_with_check

# Requires: make check (or make lint, make typecheck, make test)

Quality Commands Required:

make lint - Run linting
make typecheck - Run type checking
make test - Run tests
make check - Run all checks (or individual commands)

If these don’t exist, the workflow automatically creates a setup task and blocks until it’s resolved.

Quick Start

1. Install AnyTask CLI

# Recommended: Install script
curl -fsSL https://anyt.dev/install.sh | sh

For other installation methods, see the Installation Guide.

2. Set Up Authentication

# Set agent API key
export ANYT_API_KEY=anyt_agent_xxxxxxxxxxxxx

# Initialize workspace
anyt init --workspace-id 123 --identifier DEV

3. Get Your Agent ID

Visit https://anyt.dev/home/agents
Create a new agent if needed
Copy your agent identifier (e.g., agent-xxx)

4. Start Worker

# Start with ALL built-in workflows (recommended)
anyt worker start --agent-id agent-xxx

# Or start with specific workflow only
anyt worker start --agent-id agent-xxx --workflow local_dev

5. Create a Task

# For local_dev or dry_run (any todo task)
anyt task add "Fix typo in README" --status todo

# For feature_dev workflows (requires 'feature' label)
anyt task add "Add dark mode support" --labels feature --status todo

The worker will automatically detect the task, match it to a workflow, and execute it!

Choosing the Right Workflow

Workflow	Use Case	Quality Gates	PR Creation	Branch Management
`dry_run`	Testing/validation	No	No	No
`local_dev`	Quick iterations	Optional (continues on fail)	No	No
`feature_dev`	Standard features	No	Yes	Automatic
`feature_dev_with_check`	Production features	Yes (blocks on failure)	No	Automatic

Recommendations:

Testing setup? Use dry_run
Quick fix on current branch? Use local_dev
New feature needing PR? Use feature_dev
Production feature with quality requirements? Use feature_dev_with_check

Worker Output Example

[2025-11-15 10:00:00] Worker started
[2025-11-15 10:00:00] Agent ID: agent-xxx
[2025-11-15 10:00:00] Polling interval: 5s
[2025-11-15 10:00:00] Loaded workflows: dry_run, local_dev, feature_dev, feature_dev_with_check

[2025-11-15 10:00:01] Checking for available tasks...
[2025-11-15 10:00:02] Found task: DEV-42 (Add dark mode toggle)
[2025-11-15 10:00:02] Matched workflow: feature_dev
[2025-11-15 10:00:02] Starting attempt...

[2025-11-15 10:00:03] ✓ Started attempt #123
[2025-11-15 10:00:03] Executing workflow: Feature Development

[2025-11-15 10:00:04] Step 1/9: Create feature branch
[2025-11-15 10:00:05] ✓ Created branch: feature/DEV-42-add-dark-mode-toggle

[2025-11-15 10:00:06] Step 2/9: Analyze task requirements
[2025-11-15 10:00:15] ✓ Analysis complete

... (implementation steps) ...

[2025-11-15 10:25:30] Step 9/9: Mark task complete
[2025-11-15 10:25:31] ✓ Task marked as done

[2025-11-15 10:25:32] ✓ Workflow completed successfully
[2025-11-15 10:25:32] ✓ Finished attempt #123

[2025-11-15 10:25:33] Checking for available tasks...

Monitoring and Debugging

View Workflow Attempts

# List all attempts for a task
anyt attempt list DEV-42

# View detailed attempt information
anyt attempt show 123

# Download artifacts (like analysis, implementation summaries)
anyt artifact list 123
anyt artifact download 456 --output result.json

View Task Timeline

# See all events and updates for a task
anyt timeline DEV-42

# Filter recent events
anyt timeline DEV-42 --last 24h

Worker Logs

The worker outputs real-time logs to stdout. For production, redirect to a file or use systemd:

# Log to file
anyt worker start --agent-id agent-xxx > worker.log 2>&1

# Or use systemd (see Production Deployment below)
journalctl -u anyt-worker -f

Production Deployment

For production use, run workers as system services:

# Create systemd service file
sudo nano /etc/systemd/system/anyt-worker.service

[Unit]
Description=AnyTask Worker
After=network.target

[Service]
Type=simple
User=youruser
WorkingDirectory=/path/to/project
Environment="ANYT_API_KEY=anyt_agent_xxx"
ExecStart=/usr/local/bin/anyt worker start --agent-id agent-xxx
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

# Enable and start service
sudo systemctl enable anyt-worker
sudo systemctl start anyt-worker

# Check status
sudo systemctl status anyt-worker

# View logs
journalctl -u anyt-worker -f

Common Issues

”No workflows loaded”

Cause: Worker can’t find workflow files Solution: Built-in workflows are automatically loaded. If using custom workflows, check the path:

# List available workflows
anyt worker list-workflows

# Specify custom workflows directory
anyt worker start --agent-id agent-xxx --workflows-dir /path/to/workflows

“No tasks available”

Cause: No tasks match workflow triggers Solution: Check task status and labels:

# For local_dev or dry_run: needs status=todo
anyt task add "Test task" --status todo

# For feature workflows: needs status=todo AND feature label
anyt task add "Test feature" --status todo --labels feature

“Quality check failed” (feature_dev_with_check only)

Cause: Quality commands don’t exist or are failing Solution: The workflow automatically creates setup/fix tasks. Complete those first:

# Check what tasks were created
anyt task list --status blocked

# The workflow will have created dependency tasks - complete them first

Advanced Usage

For advanced scenarios including:

Writing custom workflows
Custom actions and steps
Complex trigger conditions
Multi-job workflows
Error handling strategies
Secrets management

See the Complete Worker Guide.

Next Steps

Quick Start

Get started with workers in 5 minutes

Complete Guide

Write custom workflows and advanced usage

Authentication

Production security best practices

CLI Reference

Worker command documentation

Support

Documentation: Complete Worker Guide
CLI Reference: Worker Commands
Email: [email protected]

Worker System

​Worker System

​Overview

​Key Features

​Built-in Workflows

​1. Dry Run (dry_run)

​2. Local Development (local_dev)

​3. Feature Development (feature_dev)

​4. Feature with Quality Control (feature_dev_with_check)

​Quick Start

​1. Install AnyTask CLI

​2. Set Up Authentication

​3. Get Your Agent ID

​4. Start Worker

​5. Create a Task

​Choosing the Right Workflow

​Worker Output Example

​Monitoring and Debugging

​View Workflow Attempts

​View Task Timeline

​Worker Logs

​Production Deployment

​Common Issues

​”No workflows loaded”

​“No tasks available”

​“Quality check failed” (feature_dev_with_check only)

​Advanced Usage

​Next Steps

Quick Start

Complete Guide

Authentication

CLI Reference

​Support

Worker System

Overview

Key Features

Built-in Workflows

1. Dry Run (`dry_run`)

2. Local Development (`local_dev`)

3. Feature Development (`feature_dev`)

4. Feature with Quality Control (`feature_dev_with_check`)

Quick Start

1. Install AnyTask CLI

2. Set Up Authentication

3. Get Your Agent ID

4. Start Worker

5. Create a Task

Choosing the Right Workflow

Worker Output Example

Monitoring and Debugging

View Workflow Attempts

View Task Timeline

Worker Logs

Production Deployment

Common Issues

”No workflows loaded”

“No tasks available”

“Quality check failed” (feature_dev_with_check only)

Advanced Usage

Next Steps

Support