ToolExperimentalMCPAG-UI

praktor

Name: praktor
Rating: 3.0 (41 reviews)
Author: mtzanidakis

by mtzanidakis

Claude-focused multi-agent orchestrator with Mission Control and Docker sandboxes

Updated Aug 14, 2026

Stars

Forks

Commits/Week

Commits/Month

View on GitHub

What It Does

Orchestrates multi-agent Claude Code workflows with Telegram I/O, Docker isolation, and a Mission Control UI. Uses swarm patterns and NATS-backed messaging to coordinate specialist agents, plus sandboxed containers for running code safely. Distinctive features include Telegram-based interaction, scheduled tasks, and a UI for supervising missions and agent activity. The design embraces a Supervisor Pattern approach to orchestrate agent coordination.

Why It Matters

As agents delegate work to other agents, operators need visibility into who did what and how failures propagate. Praktor gives a practical playground for running Claude-based agent swarms with process isolation and a central mission UI, making agent interactions auditable. That visibility helps surface agent failure modes and build early reputation signals and track records for agents in realistic, containerized runs. The concept of an Agent Card helps codify these reputational signals across runs Agent Card.

Ideal For

Developers experimenting with multi-agent Claude workflows who want a self-hosted UI, Telegram integration, and container isolation for testing agent delegation. Consider the AG-UI Protocol for standardized UI interactions, and ensure resilient coordination with the Agent-to-Agent Protocol (A2A) when delegating tasks across agents.

How It's Used

Run sandboxed agent delegations with Docker isolation to observe failure modes
Coordinate Claude-based specialists via NATS and a Mission Control UI for end-to-end debugging
Expose agent workflows to testers through Telegram for interactive, scheduled mission runs

Need implementation help?

Expert guidance available