- Category
- Cybersecurity
- Rank
- No. 2019Tools index
- Pricing
- Open Source
- Type
- AGENT
- Builder
- mksglu
- GitHub
- 32 stars
- Date
About
Secure autonomous AI agents in 60 seconds, built on OpenClaw.
What it does
SeclawAI runs a personal agent stack locally through Docker and accepts requests through Telegram. An interactive command-line setup selects an LLM provider, credentials, integrations, and agent templates. Installed capabilities can answer together through model-based routing or run alone in focus mode. The stack also supports persistent workspace data, recurring jobs, approval prompts, and OAuth-connected services.
Why it's ranked here
The appeal is concrete: runtime isolation, constrained filesystem access, outbound tunneling, scheduled work, and multiple capabilities share one conversational entry point. The implementation covers the full operating loop, from setup and health checks to tool execution and Telegram delivery. However, its security case rests largely on documented container settings not included in the supplied repository excerpts, so those guarantees deserve deployment-level verification.
What's good
Security controls are designed below the prompt layer: non-root execution, dropped Linux capabilities, read-only containers, resource limits, and a restricted workspace mount. Tool sources share one execution layer, while duplicate names are rejected. Schedules support durable steps and optional approval before delivery. Conversation history is bounded, capability responses carry attribution, and missing integrations can produce authorization links inside Telegram.
Tradeoffs
Operation requires Docker Desktop, Node.js 20 or newer, an LLM API key, and a Telegram bot token. External service access depends on Composio, while model usage remains a separate cost. Model routing decides which installed capability handles each message, so selection is probabilistic in auto mode. The agent loop stops after ten tool iterations, chat history retains twenty messages, and Telegram responses are split around its message limit.
How to use it well
It best suits technical users who want a self-hosted Telegram assistant for recurring research, inbox work, task tracking, or approval-gated automation. Start with one free capability, inspect its workspace output, then add integrations and schedules gradually. Use focus mode when routing ambiguity matters. It does not replace a general application framework, embeddable agent library, or rich browser-based chat interface.
Technical notes+
The pnpm and Turbo monorepo is declared in package.json, with Node.js 20 as the minimum engine. packages/cli/src/cli.ts defines the Commander-based setup and management surface. packages/runtime/agent.ts exposes Telegram webhook, Inngest, and health routes; packages/runtime/llm.ts implements OpenAI-compatible and direct Anthropic tool loops with a ten-iteration ceiling. packages/runtime/tools.ts combines Desktop Commander over MCP SSE, Composio REST tools, and built-ins. packages/runtime/config.ts composes installed capability prompts, while packages/runtime/scheduler.ts merges capability schedules and runs durable approval-aware jobs. packages/runtime/telegram.ts persists bounded per-chat history and handles commands, callbacks, integrations, schedules, and capability switching.
Observed
- License
- MIT
- Primary language
- TypeScript
- Packaging
- Private pnpm monorepo orchestrated with Turbo
- Install surface
- Interactive npx command-line tool; Node.js 20 or newer and Docker Desktop required
- Interfaces
- CLI management, Telegram bot messaging, MCP client tools, and internal HTTP webhook and health routes
- Platform support
- Linux, macOS, and Windows are declared by the web application metadata
- Runtime structure
- Four documented containers: agent, Inngest, Desktop Commander, and Cloudflare Tunnel
Read from README.md, package.json, packages/runtime/llm.ts, packages/runtime/agent.ts, packages/runtime/tools.ts, packages/runtime/config.ts, packages/templates/types.ts, packages/runtime/telegram.ts, packages/runtime/scheduler.ts, packages/web/tailwind.config.ts, packages/cli/src/cli.ts, packages/api/src/app.tsx, packages/web/src/app.tsx, packages/api/src/auth.ts, packages/cli/src/docker.ts.
What it can do
Deploy secure AI agents
AI agent configuration and parameters → Deployed autonomous AI agent with security protocols
Create AI agents in 60 seconds
User requirements and specifications → Functional AI agent ready for deployment
Secure AI agent communications
AI agent data and communications → Encrypted and protected agent interactions
Monitor AI agent activity
Running AI agent instances → Activity logs and performance metrics
Configure autonomous behaviors
Behavioral parameters and rules → AI agent with defined autonomous capabilities
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.