No setup required.

Bring a Clawdbotto Your Team.Instantly.

What is Clawdbot?

Clawdbot represents a fundamental shift in how we think about AI agents—from cloud-dependent services to sovereign, self-hosted systems that you truly own and control.

Clawdbot is a local-first AI agent that runs on your own machine—Mac, Linux server, NAS, or any hardware you control. Unlike cloud-based AI assistants, Clawdbot keeps your sensitive data where it belongs: on your infrastructure.

It's not just a chatbot; it's a complete agent runtime with memory, scheduled execution, and real action capabilities. Clawdbot can keep context over days and weeks, run morning briefings automatically, react to webhooks and events, and execute real operations with configurable safety rails.

Team9 makes deploying Clawdbot instant. No complex setup, no infrastructure headaches, no manual configuration of Node.js, messaging adapters, or security policies—just sovereign AI agents that work for your team, on your terms, deployed in minutes instead of days.

Whether you're automating daily reports, monitoring server health, managing your knowledge base, or orchestrating GitHub workflows, Clawdbot brings the power of AI agents to your infrastructure without sacrificing privacy or control.

Sovereign Agent Principles

Local Execution

Runs where you choose—your machine, your server, your network

Persistent Operation

Scheduled jobs, background daemons, event-driven triggers

Agency with Constraints

Real actions with safety rails and permission boundaries

Built for Control & Privacy

Data Sovereignty

Sensitive context stays local. No uploading private files, code, or credentials to opaque cloud systems.

Transparent Memory

Long-term memory stored as plain Markdown. Inspect, edit, or delete—you own your data.

Composable Tools

Extend capabilities via MCP (Model Context Protocol). Add tools without forking the core.

Why Clawdbot Went Viral

The perfect storm of technology maturity, privacy concerns, and practical utility

Data Sovereignty Pressure

Users increasingly want AI agent benefits without routinely uploading private files, internal code, calendars, or credentials into opaque cloud systems. Local-first execution addresses this anxiety directly, giving teams the confidence to use AI agents with their most sensitive data.

The "Jarvis" Moment Became Real

For years, voice assistants promised a digital assistant experience but lacked reliable action capabilities. When large language models gained tool use and planning patterns, a real action loop became practical. Clawdbot turned that possibility into reality.

Idle Hardware Found a Job

A spare Mac mini, home server, or low-power Linux box can host an always-on agent. That physical anchoring creates a sense of ownership and continuity that cloud agents rarely provide. Your agent lives where you live—on your network, your schedule, your terms.

Composable Tooling via MCP

Developers can add capabilities by attaching tool servers rather than forking a monolith. The Model Context Protocol creates a smoother ecosystem curve while letting users keep control. This composability turned Clawdbot into a platform, not just a product.

What Teams Use It For

Real-world automation that runs on your infrastructure

Daily Briefings & Reporting

Morning updates, nightly checks, and automated reports delivered to your chat channels.

Server Monitoring

Proactive health checks and instant alerts when something needs attention.

Knowledge Base Management

Summarize files, update docs, and keep your team's knowledge organized.

GitHub Operations

Automate issue triage, PR reviews, and release workflows.

How It Works

Clawdbot's architecture is modular and opinionated: Gateway handles communication, Runtime executes the plan-and-act loop, and Memory stores context transparently.

Gateway: Communication Hub

The Gateway is your agent's perimeter and communication interface. It connects to Telegram, WhatsApp, Discord, Signal, and other messaging platforms through adapters.

It handles session identity mapping, pairing policies (deny by default for unknown senders), and normalizes inbound messages before they reach the agent runtime. Think of it as the secure front door to your agent.

Agent Runtime: Planning & Action

The runtime (typically Node.js v22+) executes a reasoning-and-action loop similar to ReAct patterns: generate a plan, call tools step-by-step, feed results back to the model, and continue until complete or hitting a guardrail.

Production deployments add per-tool permission scopes, human approval for risky actions (filesystem writes, shell execution), time and cost limits, and structured logs for auditability.

Memory: Markdown Storage

Clawdbot's "Markdown memory" design is both SEO-friendly and user-friendly. Long-term memory is stored in plain text files that you can inspect, edit, or delete directly.

Deletions are real and verifiable—no hidden databases or opaque storage. Retrieval can use embeddings layered on top without hiding the underlying state. This favors transparency and reversibility, crucial for sovereign agents.

Network & Remote Access

A conservative baseline for Clawdbot deployment binds services to localhost by default and exposes remote access through an encrypted overlay network like Tailscale or a similar mesh VPN.

This approach reduces attack surface, avoids complex router port forwarding, and still allows global access to your agent from your phone or laptop. Public ports for chat gateways and tool servers are generally discouraged unless you have strong authentication and rate limiting in place.

Installation & Deployment

Get Clawdbot running on macOS, Linux, or WSL2 in minutes

Prerequisites

Node.js v22 or higher
API access to a capable LLM (tool use + planning)
2 GB+ RAM for continuous operation
Optional: Local LLM runtime (e.g., Ollama) for offline workflows

macOS

Most common path. Uses launchd for background execution.

# Install nvm

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash

# Install Node 22

nvm install 22 && nvm use 22

# Install Clawdbot

curl -fsSL https://clawd.bot/install.sh | bash

# Start daemon

clawdbot onboard --install-daemon

Linux / VPS

Production-ready deployment with security hardening.

•Run as non-root user
•Configure UFW/iptables
•Enable Fail2Ban for SSH
•Use mesh VPN (Tailscale)
•Isolate tool execution

Windows / WSL2

Recommended for Windows users via Ubuntu 24.04 LTS.

# Install WSL2

wsl --install

# Inside WSL2:

cd ~

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash

nvm install 22 && nvm use 22

⚠️Keep project inside Linux filesystem, not mounted Windows paths

Skills & Automation

Extend Clawdbot with custom workflows and scheduled operations

Skills Marketplace

Community-built capabilities for common workflows: web search, GitHub operations, note apps, ticketing systems, smart home integration, and more.

Pin versions for stability

Review code before installation

Use minimal permission scopes

Scheduled Execution

Convert your agent from reactive to proactive with cron-style schedules. Run tasks automatically without manual intervention.

08:00Morning daily briefings

*/15 *Server health checks every 15 min

22:00Nightly backup verification

Build Custom Skills

Define workflows in structured natural language using SKILL.md format. Describe inputs, outputs, and guardrails in plain text—no complex coding required.

Fast iteration—describe the workflow, test immediately, and refine based on results

Security & Risk Mitigation

Agent systems expand attack surface—here's how to stay safe

Prompt Injection

Untrusted text (webpages, docs, messages) can manipulate agent behavior.

•Isolate browsing from privileged tools
•Add approval gates for state changes
•Sandbox file access to explicit dirs

Supply Chain Attacks

Community skills and plugins introduce supply chain risk.

•Install only from verified sources
•Avoid "curl | bash" from unknown URLs
•Pin dependencies, verify checksums

Network Exposure

Publicly exposed gateways invite unauthorized access.

•Bind to localhost by default
•Use mesh VPN (Tailscale) for remote access
•Enable pairing policy for chat identities

Security Baseline

✓Per-tool permission scopes
✓Human approval for risky actions
✓Rate limiting and cost budgets

✓Structured audit logs
✓Secrets in env vars or vault
✓Container or namespace isolation

Frequently Asked Questions

Everything you need to know about Clawdbot

What is Clawdbot used for?

Daily briefings, automated reporting to chat channels, server monitoring alerts, file and note summarization, and tool-driven workflows (GitHub ops, knowledge base updates)—all running on your own machine.

Do I need a Mac mini to run Clawdbot?

No. A Mac mini is convenient for always-on hosting, but any machine you control can run Clawdbot—Mac, Linux server, NAS, or even a laptop. Availability matters more than the hardware brand.

Can Clawdbot run offline?

Yes, if you attach a local model runtime (e.g., Ollama). Some skills still require internet access (webhooks, remote APIs), but the core agent loop can be fully local.

Is Clawdbot safe?

Safety depends on deployment. A secure setup includes pairing policies, limited tool permissions, sandboxed filesystem access, approval gates for risky actions, and private network exposure through a mesh VPN. Team9 provides sensible defaults.

What is MCP in Clawdbot?

MCP (Model Context Protocol) is a way to attach external tool servers to the agent so it can access services and capabilities in a composable way. Treat MCP servers as privileged code and use allowlists and minimal permissions.

How do I install Clawdbot on Windows?

The stable route is WSL2 (Windows Subsystem for Linux) with a modern Ubuntu distribution (24.04 LTS recommended). Run all setup steps inside WSL2 and keep the project inside the Linux filesystem—avoid mounted Windows paths to prevent file locking and permission issues.

What's the difference between Clawdbot and Clawdbot?

Clawdbot is the rebranded successor to Clawdbot. The project was renamed to avoid trademark conflicts while maintaining the same core functionality and community. "Molt" refers to crustaceans shedding their shell to grow—symbolizing growth beyond constraints while keeping the lobster mascot and culture intact.

Can I use Clawdbot with local LLMs?

Yes. You can attach local model runtimes like Ollama for completely offline operation. This is ideal for privacy-sensitive workflows or air-gapped environments. Note that some skills may still require internet access for external APIs, but the core agent loop can run entirely local.

How does Team9 simplify Clawdbot deployment?

Team9 eliminates the complexity of manual setup. Instead of configuring Node.js, managing dependencies, setting up messaging adapters, and hardening security yourself, Team9 provides instant deployment with sensible defaults. You get a production-ready Clawdbot agent with zero infrastructure headaches.

Clawdbot vs Other AI Frameworks

How Clawdbot compares to AutoGPT, LangChain agents, and cloud-based alternatives

Clawdbot(Local-First Sovereign Agent)

Philosophy: Local execution, transparent memory, explicit permission boundaries. Runs on hardware you control.

Best for: Teams that prioritize data sovereignty, need persistent operation (cron jobs, event triggers), and want full control over the execution environment.

Trade-offs: Requires hosting infrastructure (Mac, Linux, VPS). Initial setup complexity reduced by Team9.

AutoGPT / AgentGPT(Autonomous Task Completion)

Philosophy: Autonomous loops that break down goals into steps and execute until complete. Focus on self-direction.

Best for: Research tasks, content generation, exploratory workflows where supervision isn't critical.

Trade-offs: Can spiral into expensive loops without strict guardrails. Less emphasis on persistent operation and chat integration.

LangChain Agents(Developer Framework)

Philosophy: Flexible building blocks for custom agent workflows. Library-first approach.

Best for: Developers building custom applications who need fine control over agent architecture and tool chains.

Trade-offs: Requires significant coding. Not an out-of-the-box agent runtime—you build your own on top of the framework.

☁️

Cloud AI Assistants(Managed Services)

Philosophy: Turnkey convenience. Zero infrastructure management. Browser or app-based access.

Best for: Individual users or teams comfortable uploading data to third-party services in exchange for ease of use.

Trade-offs: Data leaves your control. Limited customization. No persistent local operation or scheduled workflows on your infrastructure.

The Clawdbot Advantage

Clawdbot sits at the intersection of sovereignty, practicality, and ecosystem maturity. It's not just a framework (like LangChain) or a research experiment (like early AutoGPT)—it's a production-ready agent runtime that integrates with your daily tools (chat apps, GitHub, calendars) while keeping data local and operations transparent. With Team9, you get Clawdbot's power without the infrastructure burden.

Ready to Deploy

Ready to bring a sovereign AI agent to your team?

Join teams that value control, privacy, and instant deployment.