HN Showcase
Back to browse
GitHub Repository

Multi-agent Claude Code orchestrator with Telegram I/O, Docker isolation, swarm patterns, and Mission Control UI

33 starsGo

Praktor – Multi-agent Claude Code orchestrator with Docker isolation

by mtzanidakis·Feb 26, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerShip ItNiche GemBig Brain

Single Go binary: Telegram → Claude agents in isolated Docker with swarms, memory, Nix.

Strengths
  • •Docker isolation per agent eliminates blast radius—real security mindset vs. naive orchestration
  • •Swarm patterns (fan-out, pipeline, collaborative) show architectural depth beyond simple routing
  • •AES-256-GCM vault + MCP tools + hot config reload ship serious production features today
Weaknesses
  • •Requires Claude API key or OAuth token—locks users into one LLM ecosystem
  • •Early stage (2 stars)—cold start friction, unclear long-term maintenance commitment
Target Audience

AI engineers, self-hosted AI enthusiasts, teams building agent systems

Similar To

OpenClaw · AutoGen (Microsoft) · LangGraph

Post Description

I built Praktor to run my own AI agents in isolated Docker containers, accessible from Telegram.

It's a single Go binary: receives Telegram messages, routes them to named agents (each with its own role, model, and config), spins up Docker containers running Claude Code via the Agent SDK, and streams responses back. A Mission Control web UI provides real-time monitoring.

Some highlights: - Named agents with smart routing (@agent_name or AI-powered) - Each agent in its own Docker container with isolated filesystem - Persistent per-agent memory via SQLite + MCP tools - Agent swarms: fan-out, pipeline, and collaborative patterns - Encrypted secrets vault (AES-256-GCM), never exposed to the LLM - Nix package manager — agents install tools on demand - Browser automation via playwright-cli - Per-agent MCP servers, plugins, and skills from the UI - Hot config reload, scheduled tasks, backup/restore

Would love feedback.

Similar Projects

AI/ML●●Solid

A Technique for Self-Improving Agents

Git worktree isolation lets agents test instruction changes without breaking other sites—clever regression prevention.

Big BrainRabbit Hole
dataviz1000
102mo ago
AI/ML●●Solid

AgenC – an agentic work factory focused on self-upgrading

Someone actually solved the annoying UX around running dozens of agent sessions: per-agent sandboxing, a command palette with hotkeys, and built-in secrets injection make spinning up and triaging failed runs painless. There’s also an on-board assistant that can configure and drive the factory — smart move for reducing CLI friction. The tradeoff is obvious: it’s tightly coupled to macOS and Claude, so it’s brilliant if you live in that ecosystem but niche otherwise.

Rabbit HoleNiche Gem
mieubrisse
103mo ago