nrslib/takt

TAKT Agent Koordination Topology - Define how AI agents coordinate, where humans intervene, and what gets recorded — in YAML

What it solves

TAKT addresses the instability of AI coding agents by replacing unpredictable, prompt-based processes with structured, repeatable development workflows. It prevents agents from forgetting instructions, polluting context, or skipping critical review steps by treating agents as components to be orchestrated from the outside rather than trusted to manage their own process.

How it works

TAKT uses YAML-defined workflows that break a task into a sequence of explicit steps (e.g., plan → implement → review → fix). Each step is assigned a specific persona, set of permissions, and transition rules. When a task is executed, TAKT manages the transitions between these steps, ensuring that reviews cannot be skipped and that agents only receive the context necessary for their current role. Tasks are executed in isolated worktrees to keep the main codebase clean and provide a traceable path from task to pull request.

Who it’s for

It is designed for developers who use AI coding agents (such as Claude Code, Cursor, or GitHub Copilot CLI) and want to enforce a professional software development lifecycle (SDLC) without constant manual babysitting.

Highlights

  • Declarative Workflows: Define complex loops (like review-fix-re-review) in YAML files.
  • Provider Agnostic: Supports a wide range of agents including Claude Code, DeepSeek Harness, Cursor, and GitHub Copilot CLI.
  • Isolated Execution: Runs tasks in separate worktrees by default to prevent working-tree pollution.
  • Role Separation: Assigns distinct personas and output contracts to different steps to maintain focus and reduce context bloat.
  • Integration: Connects with GitHub and GitLab issues to turn tickets into queued AI tasks.

Related

  • Project
    whitelonng/mancode

    An AI coding agent workflow orchestration tool that adds a layer of engineering discipline, structured planning, and cross-session continuity to existing AI agents.

  • Project
    coleam00/Archon

    A workflow engine for AI coding agents that uses YAML definitions to make software development processes deterministic, repeatable, and isolated.

  • Project
    rpamis/comet

    A resumable long-running task workflow and skill platform for AI-assisted coding that provides structured methodologies and multi-agent coordination for complex software requirements.

  • Project
    AnandChowdhary/continuous-claude

    A CLI tool that runs AI coding agents in a continuous loop to autonomously create, test, and merge pull requests for large-scale software projects.

  • Project
    soumatheusgomes/vibe-coding-toolkit

    A documented workflow and set of tools (plugins, prompts, lint rules, sub‑agent orchestration, memory handling) that let developers use Claude Code / Codex as a disciplined coding teammate. Includes a “Superpowers” discipline, parallel sub‑agents, quality‑gate ESLint config, and a full onboarding Playbook.