seclaw

Agentic AI on your machine. Multiple agents, one Telegram bot.
Self-hosted. Docker-isolated. Secure by default.

Quick Start · Auto Mode · Templates · Architecture · Security

npx seclaw

One command. Pick your LLM, enter your Telegram token, choose a template. Agent live in 60 seconds.

OpenClaw proved the demand — 68K+ GitHub stars, 600K people watched the tutorial. But it ships with zero container isolation: API keys exposed to every MCP container, your home directory mounted to root-privileged processes, port 5678 open to the internet with no auth.

seclaw fixes both problems. Same autonomous agent capabilities — with Docker-level security enforced at runtime, not in the system prompt. One can be jailbroken. The other can't.

Prerequisites: Docker Desktop, Node.js 20+, an LLM API key (OpenRouter recommended / Anthropic / OpenAI / Google), a Telegram bot token.

npx seclaw

The CLI handles everything: LLM provider, API key, Telegram token, Composio integrations, template selection, docker compose up. No config files. No YAML. No Docker knowledge required.

npx seclaw status    # Check all services
npx seclaw doctor    # Auto-diagnose & fix issues

Install multiple agents and they work together automatically. No manual switching — the LLM routes each message to the right capability.

You: "Summarize my unread emails and check HN for AI news"

seclaw:
  Inbox: 3 urgent, 7 action needed, 12 FYI.
         Top urgent: contract review from legal (deadline tomorrow).

  Hacker News: Top AI stories — Anthropic launches Claude 4.5,
               Google open-sources Gemma 3, OpenAI restructures.

  — Inbox Management, Research & Intelligence

Every response ends with an attribution footer showing which agent(s) answered.

1. Install agents — npx seclaw add inbox-agent --key TOKEN. Each template stacks onto your base agent.
2. Auto-route — The LLM picks the right capability for each message. Email questions go to Inbox. Lead questions go to Sales.
3. Attribution — Every response shows which agent answered. You always know who did what.

Use /templates in Telegram or npx seclaw templates in CLI to view and switch modes. Auto mode is the default.

/templates           → List installed templates + switch mode

npx seclaw templates  # List installed templates + active mode

When an agent needs an integration that isn't connected, it generates an OAuth link directly in chat. No manual setup needed.

seclaw: I need X (Twitter) connected to scan for leads.
        Authorize X (Twitter): https://composio.dev/auth/...

        Open the link, sign in, and grant access.
        After completing authorization, your new tools load automatically.

The connect_integration tool handles this programmatically — the agent calls it when it detects a missing integration, gets the OAuth URL, and shares it with the user.

17 agent templates. Self-hosted. No subscriptions. Your data stays on your machine.

All templates work in Auto Mode — install multiple and they collaborate. No conflicts, no switching.

When you purchase a template, you get a single token that never expires. Every future purchase is added to the same token — no separate keys, no expiry, re-download anytime.

# Free templates — no token needed
npx seclaw add productivity-agent

# Paid templates — use your token
npx seclaw add inbox-agent --key YOUR_TOKEN
npx seclaw add research-agent --key YOUR_TOKEN   # same token
npx seclaw status                                  # Check services & installed

Browse and purchase at seclawai.com/templates.

4 Docker containers, zero inbound ports, full isolation:

Every guardrail is enforced at the Docker runtime level — not in the system prompt.

16 tools built into every agent — no MCP dependency, no external services. All available via natural language in Telegram.

Every agent gets a structured workspace at /workspace:

/workspace
├── memory/       Persistent learnings about you (auto-updated)
├── tasks/        TODOs and action items with priority + due dates
├── notes/        Quick thoughts, ideas, meeting notes, links
├── reports/      Research results, summaries, daily digests
├── drafts/       Draft emails, messages, documents to review
└── config/       Schedules, templates, integrations (system-managed)

When an agent needs Twitter, Reddit, YouTube, or any other service — it calls connect_integration, gets an OAuth URL, and sends it to the user. No manual /integrations needed.

Powered by Composio — managed OAuth, zero raw token storage.

Gmail, Google Calendar, Google Drive, Google Sheets, GitHub, Slack, Notion, Linear, Trello, Todoist, Dropbox, WhatsApp, X (Twitter), Reddit, YouTube, Tavily, Hacker News.

npx seclaw integrations    # CLI: connect/disconnect
/integrations              # Telegram: inline buttons

Or let the agent handle it — when a template needs an integration, it calls connect_integration and sends you the OAuth link directly.

npx seclaw                           # Interactive setup wizard
npx seclaw add <template> --key TOKEN # Download a purchased template
npx seclaw templates                 # List installed templates & mode
npx seclaw integrations              # Connect/disconnect services
npx seclaw status                    # Check services & health
npx seclaw doctor                    # Auto-diagnose & fix issues
npx seclaw stop                      # Stop all containers
npx seclaw upgrade                   # Pull latest images & restart

Your only real cost is the LLM API. Everything else is free — Docker, Inngest, Cloudflare Tunnel, Telegram, Composio free tier.

git clone https://github.com/seclaw/seclaw.git && cd seclaw
pnpm install          # auto-builds CLI
npx seclaw            # local version, always up-to-date

MIT

Website · Templates · GitHub