task-runner

Personal Task Execution Agent

A self-healing LangGraph agent that accepts high-level personal tasks, breaks them into steps, executes them, validates results, and loops until completion — with full observability in LangSmith.

Now with OpenClaw integration - Run your agent 24/7 and access it from WhatsApp, Telegram, Discord, Slack, and more!

Architecture

START
  │
  ▼
┌─────────┐   LLM decomposes task into ordered steps
│ PLANNER │──────────────────────────────────────────►┐
└─────────┘                                           │
                                                      ▼
                                               ┌──────────────┐
                                      ◄────────│   EXECUTOR   │
                                      │        └──────┬───────┘
                                      │               │ all steps done
                                      │               ▼
                                      │        ┌──────────────┐
                    re-plan after fix │        │  VALIDATOR   │
                                      │        └──────┬───────┘
                                      │               │ issues found
                                      │               ▼
                                      └────────┌──────────────┐
                                               │    FIXER     │
                                               └──────────────┘
                                                      │ passed
                                                      ▼
                                               ┌──────────────┐
                                               │   REPORTER   │──► END
                                               └──────────────┘

Tools

| Tool | What it does | |------|-------------| | file_operations | Read, write, copy, move, delete files | | folder_operations | Create, list, organize folders | | web_search | Search the web for information | | web_fetch | Fetch content from URLs | | send_email | Send emails (stub - wire to SMTP) | | calendar_event | Create calendar events (stub) | | shell_command | Run arbitrary shell commands | | data_to_excel | Convert data to Excel format |

Quick start

Option 1: CLI Mode (Local)

# 1. Install dependencies
pip install -r requirements.txt

# 2. Configure env vars
cp .env.example .env
# Edit .env and fill in your LANGSMITH_API_KEY and OPENAI_API_KEY

# 3. Run
python main.py

Option 2: OpenClaw Mode (24/7 with Messaging)

# 1. Install OpenClaw
curl -fsSL https://openclaw.ai/install.sh | bash

# 2. Install Python dependencies
pip install -r requirements.txt

# 3. Install Node dependencies
npm install

# 4. Configure environment
cp .env.example .env
# Edit .env with your API keys

# 5. Link plugin to OpenClaw
openclaw plugin link .

# 6. Start OpenClaw gateway
openclaw gateway start

# 7. Configure your messaging channels
openclaw channel add telegram  # or discord, whatsapp, slack

Now you can message your agent from any connected platform!

Example messages:

• "Organize my downloads folder by file type"
• "Task: Create weekly report | Outcome: Excel file with charts"
• "/status" - Check if agent is online
• "/help" - Show usage instructions

Example tasks

• "Organize my downloads folder by file type"
• "Search for Python best practices and create a summary document"
• "Create a weekly report from CSV files in the data folder"
• "Send reminder emails to my team about the meeting"