OpenXJarvis (openclaw-python)

Jarvis lives here. A pure Python clone of OpenClaw — the most powerful personal AI gateway you can run.

Keywords: AI agent framework · autonomous agent · LLM gateway · multi-channel chatbot · self-hosted AI assistant · Python AI · Telegram bot · Discord bot · personal AI · agent orchestration

What is This?

OpenXJarvis is a complete Python implementation of OpenClaw — continuously aligned with the TypeScript reference. Think of it as your personal AI operating system: an autonomous agent platform that connects messaging channels (Telegram, WhatsApp, Discord, Slack, IRC, Signal, Google Chat, iMessage), integrates with cutting-edge LLM providers (OpenAI GPT-5.2/5.3, Claude Opus 4.5/4.6, Google Gemini 3.x/2.5, AWS Bedrock, Ollama), manages persistent memory and long-term context, schedules autonomous tasks via cron, and provides a beautiful Web UI for agent orchestration.

Why "Jarvis"? Because this isn't just another chatbot framework or AI assistant wrapper. This is a production-grade agent framework — a self-hosted personal AI that learns, remembers, and operates autonomously. Every feature, every alignment improvement, every new capability lands here.

This project stays synchronized with OpenClaw. We track upstream changes and continuously merge improvements. You're not choosing between Python and TypeScript — you're choosing the same powerful platform in your preferred language.

Tested Configurations

Component	Status	Notes
Telegram + Gemini	✅ Production Ready	Fully tested and operational
WhatsApp	🧪 Verification in progress	QR link integration complete
Discord	🧪 Verification in progress	Bot API integration complete
Slack	🧪 Verification in progress	Socket Mode integration complete
IRC	🧪 Verification in progress	Classic IRC networks support
GPT-5.2 Pro	🧪 Verification in progress	OpenAI API integration complete
Claude Opus 4.6	🧪 Verification in progress	Anthropic API integration complete
Gemini 3.1 Pro	🧪 Verification in progress	Google API integration complete
Grok 4	🧪 Verification in progress	xAI API integration complete
DeepSeek V3.2	🧪 Verification in progress	DeepSeek API integration complete
Web UI	✅ Production Ready	Full feature parity with TypeScript
Cron Scheduler	✅ Production Ready	Autonomous task execution working

Latest test: Telegram bot + Gemini 3 Pro Preview running complex multi-tool workflows (web search, file operations, reasoning) — rock solid. 🚀

Model Updates (Feb 2026): Full support for GPT-5.2/5.3 Codex, Claude Opus 4.5/4.6, Gemini 3 Pro/Flash Preview, Gemini 2.5 Pro/Flash. Additional providers (Zhipu GLM-5, xAI Grok, DeepSeek, Qwen) coming soon.

Preview

Jarvis responding on Telegram — powered by OpenXJarvis

Dependencies: pi-mono-python

openclaw-python depends on pi-mono-python — a companion repo that provides the core agent and LLM infrastructure as local packages:

Package	Provides
`pi-ai`	Unified LLM streaming layer (Gemini, Anthropic, OpenAI, …)
`pi-agent`	Agent loop, tool execution, session state
`pi-coding-agent`	Coding agent with file/bash/search tools
`pi-tui`	Terminal UI rendering engine

These packages are resolved by uv as path dependencies relative to the parent directory. Both repos must be cloned as siblings inside the same parent directory (the name of the parent directory does not matter):

my-workspace/              ← any name works
├── openclaw-python/       ← this repo
└── pi-mono-python/        ← required sibling

Getting Started

1. Install

Prerequisites: Python 3.11+ and uv package manager.

# Install uv (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Create a workspace directory (any name) and clone both repos into it
mkdir my-workspace && cd my-workspace

git clone https://github.com/openxjarvis/pi-mono-python.git
git clone https://github.com/openxjarvis/openclaw-python.git

# Install (uv resolves pi-mono-python packages automatically)
cd openclaw-python
uv sync

2. Onboard

Run the interactive setup wizard (once per environment):

uv run openclaw onboard

The wizard walks you through:

LLM provider and model selection (Anthropic, OpenAI, Gemini, Ollama)
Optional fallback models for resilience
Telegram bot setup
Gateway port and authentication
Workspace and personality initialization

It prompts for API keys interactively and saves them to .env automatically — no need to create .env by hand first.

If you already have keys in .env, the wizard detects and reuses them:

# Optional: pre-populate before onboarding
cp .env.example .env
# edit .env with your keys, then run the wizard

Onboarding creates ~/.openclaw/ with your config and workspace. Run this once per new environment.

3. Start

uv run openclaw start

That's it. The gateway and all configured channels start in the foreground. Open http://localhost:18789 to access the Web UI.

Starting the Gateway

The gateway is the core server. It manages the agent runtime, channels, sessions, cron, and serves the Web Control UI.

Quick summary: Use uv run openclaw start for the simplest one-command startup. Use uv run openclaw gateway install if you want it to run automatically as a system service.

Option A — One-command start (simplest)

uv run openclaw start

Starts everything (gateway + Telegram channel) in the foreground. Logs print to the terminal. Press Ctrl+C to stop. This is the easiest way to get running. Use it for development or quick local use.

Option B — Foreground gateway only

uv run openclaw gateway run

Same as above but starts only the gateway server (no channels). Useful when you want to control channels separately.

Option C — Background daemon (recommended for always-on use)

If you want the gateway to start automatically and keep running in the background (even after a reboot), install it as a system service first:

# 1. Install as a system service (runs once — launchd on macOS, systemd on Linux)
uv run openclaw gateway install

# 2. Then manage it with:
uv run openclaw gateway start       # start the daemon
uv run openclaw gateway stop        # stop it
uv run openclaw gateway restart     # restart
uv run openclaw gateway status      # check if it's running
uv run openclaw gateway logs        # tail the log file

Important: uv run openclaw gateway install and uv run openclaw start are not interchangeable:

Command What it does

uv run openclaw start Runs directly in the terminal (foreground). Stops when you close the terminal.

uv run openclaw gateway install Registers a background system service. Survives terminal close and reboots. Requires running gateway start after install.

A common mistake is running gateway install and then also start at the same time — this would launch two instances that both try to bind to the same port. Pick one approach and stick with it.

Web Control UI

Once the gateway is running, open your browser at:

http://localhost:18789

The UI lets you chat with the agent, inspect sessions, manage cron jobs, and configure settings.

CLI Commands

Core Commands

Command	Description
`uv run openclaw start`	Start gateway + channels in the foreground
`uv run openclaw onboard`	Interactive setup wizard (run once)
`uv run openclaw doctor`	Run system diagnostics
`uv run openclaw version`	Show OpenClaw version
`uv run openclaw tui`	Launch Terminal UI

Gateway Management

Command	Description
`uv run openclaw gateway run`	Start gateway only in the foreground
`uv run openclaw gateway install`	Install as a background system service (one-time setup)
`uv run openclaw gateway start`	Start the installed background service
`uv run openclaw gateway stop`	Stop the background service
`uv run openclaw gateway restart`	Restart the background service
`uv run openclaw gateway status`	Check background service status
`uv run openclaw gateway logs`	Tail gateway log file
`uv run openclaw gateway uninstall`	Remove the background service

Configuration

Command	Description
`uv run openclaw config show`	Show current configuration
`uv run openclaw config get <key>`	Get a config value
`uv run openclaw config set <key> <value>`	Set a config value
`uv run openclaw config unset <key>`	Remove a config value
`uv run openclaw directory`	Show OpenClaw directories

Models

Command	Description
`uv run openclaw models status`	Show configured model and fallbacks
`uv run openclaw models set <model>`	Set the default model
`uv run openclaw models fallbacks list`	List fallback models
`uv run openclaw models fallbacks add <model>`	Add a fallback model
`uv run openclaw models fallbacks remove <model>`	Remove a fallback model

Channels & Pairing

Command	Description
`uv run openclaw channels list`	List configured channels
`uv run openclaw channels status`	Show channel status
`uv run openclaw pairing list <channel>`	List pending pairing requests
`uv run openclaw pairing approve <channel> <code>`	Approve a pairing request
`uv run openclaw pairing deny <channel> <code>`	Deny a pairing request
`uv run openclaw pairing clear <channel>`	Clear all pending pairing requests
`uv run openclaw pairing allowlist <channel>`	Show allowFrom list for a channel

Cron (Scheduled Tasks)

Command	Description
`uv run openclaw cron list`	List scheduled jobs
`uv run openclaw cron add`	Add a new cron job (interactive)
`uv run openclaw cron run <job-id>`	Force-run a job immediately
`uv run openclaw cron remove <job-id>`	Remove a cron job
`uv run openclaw cron enable <job-id>`	Enable a cron job
`uv run openclaw cron disable <job-id>`	Disable a cron job

Agent & Sessions

Command	Description
`uv run openclaw agent run`	Run an agent turn via the Gateway
`uv run openclaw agent agents`	Manage isolated agents
`uv run openclaw message send <channel> <target>`	Send a message to a channel
`uv run openclaw memory search <query>`	Search memory index
`uv run openclaw memory rebuild`	Rebuild memory index

Skills & Tools

Command	Description
`uv run openclaw skills list`	List available skills
`uv run openclaw skills show <name>`	Show skill details
`uv run openclaw skills refresh`	Refresh skills cache
`uv run openclaw tools list`	List available tools
`uv run openclaw tools show <name>`	Show tool details

Cleanup & Maintenance

Command	Description
`uv run openclaw cleanup`	Clean up processes, ports, and stale state
`uv run openclaw cleanup --kill-all`	Kill all openclaw processes
`uv run openclaw cleanup --ports 18789`	Free a specific port
`uv run openclaw cleanup --stale`	Remove stale lock/state files only
`uv run openclaw logs tail`	Tail gateway log file
`uv run openclaw logs clear`	Clear log files

Advanced

Command	Description
`uv run openclaw approvals list`	List pending approvals
`uv run openclaw acp`	Launch Approvals Control Panel
`uv run openclaw hooks list`	List lifecycle hooks
`uv run openclaw plugins list`	List loaded plugins
`uv run openclaw security permissions`	Show security permissions
`uv run openclaw system heartbeat`	Trigger system heartbeat
`uv run openclaw browser launch`	Launch OpenClaw dedicated browser

Telegram Setup

Message @BotFather on Telegram → /newbot → copy the token
Add TELEGRAM_BOT_TOKEN=<your-token> to .env
Start the gateway: uv run openclaw gateway run
Open your bot on Telegram and start chatting

Access control: By default, new users must be approved. Use /pair in the bot to request access, then approve from the CLI:

uv run openclaw pairing list telegram
uv run openclaw pairing approve telegram <code>

Session commands (in Telegram chat):

Command	Action
`/reset`	Start a fresh conversation session
`/cron`	View and manage scheduled tasks
`/help`	Show available commands

Architecture Overview

openclaw-python/
├── openclaw/
│   ├── agents/              # Agent runtime, session management, context
│   │   ├── providers/       # LLM providers (Anthropic, Gemini, OpenAI)
│   │   ├── tools/           # 24+ built-in tools
│   │   └── skills/          # 56+ modular skills (loaded at runtime)
│   ├── channels/            # Messaging integrations
│   │   └── telegram/        # Telegram channel (fully operational)
│   ├── gateway/             # WebSocket gateway server
│   │   ├── api/             # RPC method handlers
│   │   └── protocol/        # WebSocket frame protocol
│   ├── cron/                # Cron scheduler (job store, timer, execution)
│   ├── infra/               # System events, in-memory queues
│   ├── config/              # Configuration loading and schema
│   ├── routing/             # Session key resolution and routing
│   └── cli/                 # Command-line interface
└── tests/                   # Unit and integration tests

Data directory (~/.openclaw/):

~/.openclaw/
├── openclaw.json            # Gateway configuration
├── workspace/               # Agent workspace (injected into system prompt)
│   ├── SOUL.md              # Personality and values
│   ├── AGENTS.md            # Operating instructions
│   ├── TOOLS.md             # Tool configurations
│   └── USER.md              # User profile
├── agents/main/sessions/    # Conversation sessions (UUID-based)
├── cron/                    # Cron job store and logs
└── logs/                    # Gateway and channel logs

Cron (Scheduled Tasks)

The cron system lets you schedule one-shot or recurring tasks that fire agent turns automatically.

# List scheduled jobs
uv run openclaw cron list

# Add a one-shot job (via CLI)
uv run openclaw cron add --name "Daily check" --schedule "0 9 * * *"

# Force-run a job immediately
uv run openclaw cron run <job-id>

You can also manage cron jobs from the Web UI or Telegram via /cron.

Cron events are delivered to the active session and optionally forwarded to Telegram.

Skills and Tools

Skills are markdown + optional Python files loaded from ~/.openclaw/workspace/skills/ or the built-in skills directory. They extend the agent's knowledge and capabilities without modifying core code.

Tools are Python callables available to the agent at runtime. Built-in tools include file operations, web search, code execution, system queries, memory management, and more.

Access Control

Edit ~/.openclaw/openclaw.json to configure the DM policy for Telegram:

{
  "channels": {
    "telegram": {
      "dm_policy": "pairing"
    }
  }
}

Policies:

Policy	Behavior
`pairing` (default)	New users must request and be approved
`allowlist`	Only pre-approved users can interact
`open`	Any user can interact (use with caution)
`disabled`	No DM access

Development

# Run test suite
uv run pytest

# Run specific tests
uv run pytest tests/integration/

# Lint and format
uv run ruff check .
uv run ruff format .

# Build Web UI (if modifying frontend)
cd openclaw/web/ui-src
npm install && npm run build

Development Status

Core Infrastructure

Component	Status	TypeScript Alignment
Gateway server	✅ Production	100%
Session management	✅ Production	100%
Context compaction	✅ Production	100%
Tool system (24+ tools)	✅ Production	100%
Skill system (56+ skills)	✅ Production	100%
Cron scheduler	✅ Production	100%
Memory indexing	✅ Production	100%
Web Control UI	✅ Production	100%

Channels

Channel	Status	Verification
Telegram	✅ Production	✅ Battle-tested
Discord	✅ Implemented	🧪 In verification
Slack	✅ Implemented	🧪 In verification
WhatsApp	✅ Implemented	🧪 In verification
IRC	✅ Implemented	🧪 In verification

AI Models

Provider	Status	Verification
Google Gemini	✅ Production	✅ Battle-tested
Anthropic Claude	✅ Implemented	🧪 In verification
OpenAI GPT	✅ Implemented	🧪 In verification
Local LLMs (Ollama)	✅ Implemented	🧪 In verification

Roadmap

🎯 Next: Complete verification of all channels and models
🎯 Q2 2026: Voice integration
🎯 Continuous: Maintain 100% alignment with OpenClaw TypeScript

Mission

We're building the future of personal AI — in Python.

OpenClaw showed the world what's possible: a personal AI that doesn't just chat, but operates. It schedules tasks, manages memory, coordinates across channels, and grows with you. Now we're bringing that power to the Python ecosystem.

This isn't a fork that will drift. This isn't a "spiritual successor." This is a living clone — continuously synchronized, feature-for-feature, with OpenClaw's evolution. Every breakthrough, every optimization, every new capability: we bring it here.

Why does this matter?

Python's ML/AI ecosystem is unmatched
Easier deployment for data scientists and ML engineers
Same powerful architecture, Python-native performance
Your agent's capabilities grow as OpenClaw grows

Join us. Whether you're building your own Jarvis, experimenting with autonomous agents, or just want a badass AI assistant — this is your platform.

OpenClaw TypeScript Feature Parity

This Python implementation tracks OpenClaw TypeScript feature-for-feature. Below is the complete list of channels and AI providers supported by OpenClaw TS — our roadmap for full parity.

Messaging Channels (TS Support)

Channel	Status in Python	Notes
Telegram	✅ Production	Bot API — fully tested
WhatsApp	✅ Implemented	QR link — verification in progress
Discord	✅ Implemented	Bot API — verification in progress
Slack	✅ Implemented	Socket Mode — verification in progress
IRC	✅ Implemented	Classic IRC networks with pairing
Google Chat	📋 Planned	Google Workspace Chat app
Signal	📋 Planned	signal-cli linked device
iMessage	📋 Planned	macOS-only, work in progress in TS

AI Model Providers (TS Support)

Provider	Status in Python	Latest Models (Feb 2026)
OpenAI	✅ Implemented	GPT-5.2, GPT-5.3 Codex, GPT-5 Mini, GPT-4o, o3-mini
Anthropic	✅ Implemented	Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4, Claude 3.7 Sonnet, Claude 3.5 Haiku
Google Gemini	✅ Implemented	Gemini 3 Pro/Flash Preview (Feb 2026 🆕), Gemini 2.5 Pro/Flash, Gemini 2.0 Flash
AWS Bedrock	✅ Implemented	Claude 3.x, Llama 3.3, Mistral, Command R+, Titan
Ollama	✅ Implemented	Llama 3.3, DeepSeek-Coder, Mistral, CodeLlama, Qwen (local)
xAI (Grok)	🚧 Q2 2026	Grok 4, Grok 3
Zhipu AI	🚧 Q2 2026	GLM-5 (745B MoE), GLM-4.7
DeepSeek	🚧 Q2 2026	DeepSeek V3.2 (671B MoE), DeepSeek-R1
Alibaba	🚧 Q2 2026	Qwen 3.5 Max, QwQ-32B-Preview
Together AI	📋 Future	100+ open-source models
Hugging Face	📋 Future	Inference API models
Replicate	📋 Future	Cloud-hosted models
Groq	📋 Future	Ultra-fast inference
MiniMax	📋 Future	MiniMax-M2.1, MiniMax-VL-01
Moonshot	📋 Future	Moonshot AI models

Recommended Models (Feb 2026):

🏆 GPT-5.2 — Best overall performance
💻 Claude Opus 4.6 / GPT-5.3 Codex — Top coding assistants
💰 Gemini 3 Flash / Gemini 2.5 Flash — Best value
🧠 Claude Opus 4.5 / Gemini 3 Pro — Advanced reasoning
🚀 Gemini 3 Pro/Flash Preview — Latest from Google (Feb 2026 🆕)

API Formats Supported

✅ OpenAI Completions API (ChatGPT, compatible endpoints)
✅ Anthropic Messages API (Claude, compatible endpoints)
✅ Google Generative AI (Gemini native)
✅ Ollama Native API
📋 OpenAI Responses API (Realtime/streaming)

Voice & Media (TS Support)

Feature	Status in Python	Notes
Voice input (STT)	📋 Planned	OpenAI Whisper, Google STT
Voice output (TTS)	📋 Planned	OpenAI TTS, Google TTS, ElevenLabs
Image understanding	✅ Implemented	Via vision models (GPT-4V, Gemini Vision, Claude)
Video understanding	📋 Planned	Gemini video analysis
Audio transcription	📋 Planned	Whisper integration

Related Projects & Ecosystem

Core Infrastructure

OpenClaw TypeScript: github.com/openclaw/openclaw — upstream reference implementation
pi-mono-python: github.com/openxjarvis/pi-mono-python — core agent infrastructure

Agent Skills Collections

Awesome Claude Skills: github.com/ComposioHQ/awesome-claude-skills — curated Claude skills with app automation (38k+ stars)
Antigravity Awesome Skills: github.com/sickn33/antigravity-awesome-skills — 950+ agentic skills for Claude Code, Gemini CLI, Cursor (16k+ stars)
Awesome Agent Skills: github.com/VoltAgent/awesome-agent-skills — 380+ skills from official dev teams (Anthropic, Google, Vercel, Microsoft) (8.5k+ stars)
Awesome OpenClaw Skills: github.com/VoltAgent/awesome-openclaw-skills — 2,868+ OpenClaw skills from ClawHub registry (21k+ stars)

Community

Join the discussion for AI agents, LLM orchestration, multi-channel bots, and autonomous AI development

Topics

python ai-agent llm chatbot telegram-bot discord-bot autonomous-agent ai-assistant openai-api anthropic-claude google-gemini self-hosted websocket cron-scheduler agent-framework ai-gateway multi-channel session-management memory-management agent-orchestration

Remote Repository

https://github.com/openxjarvis/openclaw-python

Upstream: https://github.com/openclaw/openclaw (TypeScript reference)

License

MIT — see LICENSE for details.

Name		Name	Last commit message	Last commit date
Latest commit History 175 Commits
.github		.github
assets		assets
extensions		extensions
git-hooks		git-hooks
openclaw		openclaw
scripts		scripts
skills		skills
tests		tests
workspace		workspace
.dockerignore		.dockerignore
.env.example		.env.example
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
CONTRIBUTORS.md		CONTRIBUTORS.md
Dockerfile		Dockerfile
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
VERSION.txt		VERSION.txt
docker-compose.yml		docker-compose.yml
pairing		pairing
pyproject.toml		pyproject.toml
pytest.ini		pytest.ini
start_clean.sh		start_clean.sh
uv.lock		uv.lock

Command	What it does
`uv run openclaw start`	Runs directly in the terminal (foreground). Stops when you close the terminal.
`uv run openclaw gateway install`	Registers a background system service. Survives terminal close and reboots. Requires running `gateway start` after install.

Uh oh!

License

openxjarvis/openclaw-python

Folders and files

Latest commit

History

Repository files navigation

OpenXJarvis (openclaw-python)

What is This?

Tested Configurations

Preview

Dependencies: pi-mono-python

Getting Started

1. Install

2. Onboard

3. Start

Starting the Gateway

Option A — One-command start (simplest)

Option B — Foreground gateway only

Option C — Background daemon (recommended for always-on use)

Web Control UI

CLI Commands

Core Commands

Gateway Management

Configuration

Models

Channels & Pairing

Cron (Scheduled Tasks)

Agent & Sessions

Skills & Tools

Cleanup & Maintenance

Advanced

Telegram Setup

Architecture Overview

Cron (Scheduled Tasks)

Skills and Tools

Access Control

Development

Development Status

Core Infrastructure

Channels

AI Models

Roadmap

Mission

OpenClaw TypeScript Feature Parity

Messaging Channels (TS Support)

AI Model Providers (TS Support)

API Formats Supported

Voice & Media (TS Support)

Related Projects & Ecosystem

Core Infrastructure

Agent Skills Collections

Community

Topics

Remote Repository

License

About

Resources

License

Contributing

Uh oh!

Stars

Watchers

Forks

Releases

Sponsor this project

Uh oh!

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages