TTal CLI: Orchestrate Multiple Claude Code Agents for Autonomous PR Workflows

TTal CLI: Orchestrate Multiple Claude Code Agents for Autonomous PR Workflows

TTal is a Go CLI that creates a multi-agent system with persistent manager agents and disposable worker agents, letting you run entire PR cycles from your phone via Telegram.

2h ago·3 min read·10 views·via hn_claude_code
Share:

TTal CLI: Orchestrate Multiple Claude Code Agents for Autonomous PR Workflows

What It Does

TTal is a command-line tool built in Go that transforms Claude Code from a single-session assistant into a multi-agent software factory. It solves the "babysitting" problem developers face when managing complex PR workflows—where you constantly switch windows, copy-paste feedback, and direct Claude through iterative fixes.

The architecture uses two distinct planes:

Manager Plane: Long-running Claude Code agents that persist across sessions. These agents handle high-level planning, break projects into tasks, assign priorities, and unblock workers when they get stuck. You interact with them through Telegram, allowing management from anywhere—including your phone.

Worker Plane: Short-lived agents spawned per specific task. Each worker gets an isolated git worktree and runs in its own tmux session. Workers handle implementation, code review, feedback triage, and merging. The entire PR loop runs autonomously, with CI logs fed back into the system. When tasks complete, everything cleans up automatically.

All components—manager agents, worker agents, and humans—communicate through a peer-to-peer message bridge. Any agent can message any other agent or human directly, creating a flexible collaboration system.

Setup

Installation is straightforward via Homebrew:

brew tap tta-lab/ttal
brew install ttal

The system is intentionally composable. It uses:

  • Taskwarrior for task management
  • Flicknote or plain markdown files for knowledge/memory persistence
  • Telegram as the human-in-the-loop interface

Everything communicates via CLI commands—no SDK required. This makes it easy to integrate with existing workflows and tools.

When To Use It

TTal shines in several specific scenarios:

Complex PR Management: When you have multiple PRs in flight with iterative feedback cycles. Instead of manually guiding Claude through each review-comment-fix cycle, TTal workers handle the entire loop autonomously.

Long-Term Project Planning: Need dedicated agents for ongoing research or architectural planning? TTal's manager plane provides persistent Claude Code sessions that maintain context across days or weeks, unlike typical throwaway coding sessions.

Parallel Development: Spawn multiple worker agents to tackle different tasks simultaneously. Each gets its own isolated environment, preventing conflicts while maintaining progress visibility through the message bridge.

Mobile Development Workflow: Manage your development process from Telegram while away from your computer. Review agent plans, approve next steps, or provide guidance—all from your phone.

Related Tool: Logos

The creator also built Logos, a bash-only subagent runtime with zero tool call overhead. This complements TTal by providing lightweight agent execution when you don't need full Claude Code capabilities. Available at: https://github.com/tta-lab/logos

Architecture Benefits

This two-plane approach separates concerns effectively:

  • Managers think strategically and maintain long-term context
  • Workers execute tactically in disposable environments
  • Humans oversee through high-level messaging, not micromanagement

The p2p message bridge ensures information flows where needed without centralized bottlenecks. Isolated git worktrees prevent workers from interfering with each other or your main codebase.

Getting Started

After installation, configure your Telegram integration and set up your manager agents. Start with a simple project to understand the workflow:

  1. Have a manager agent analyze a GitHub issue
  2. Watch it break the work into tasks
  3. Observe workers spawn, implement, and handle PR feedback
  4. Intervene via Telegram only when necessary

This system represents a significant evolution in how developers can leverage Claude Code—from interactive assistant to autonomous engineering team.

AI Analysis

Claude Code users should start thinking about agent specialization. Instead of using Claude Code as a general-purpose tool for everything, consider creating dedicated agents for specific roles: one for planning/architecture, another for implementation, a third for code review. Set up TTal for any project with multiple interdependent tasks. The manager-worker separation means you can have a persistent "architect" agent that understands the big picture while disposable "implementer" agents handle the grunt work. This preserves context window efficiency—managers don't waste tokens on implementation details, workers don't waste tokens on architectural reasoning. Use the Telegram interface to stay in the loop without being tied to your terminal. This changes Claude Code from something you actively use to something that works for you in the background. For teams, consider having different manager agents for different domains (frontend, backend, DevOps) that can coordinate through the message bridge.
Original sourcenews.ycombinator.com

Trending Now

More in Products & Launches

Browse more AI articles