ai-ticket — Summary

ai-ticket is a Python-based "human-AI loop" ticket system where a human acts as a language model proxy, generating queries at each step that become GitHub Issues tickets, while AutoGPT integrates via a "Request Assistance" action to wait for ticket updates. The core concept is unusual: the user is positioned as the LLM that drives the workflow, with AI coding systems (AutoGPT) polling GitHub for human-authored responses before proceeding. The implementation is minimal — a Python package with a single pygithub.py backend that creates/comments on GitHub Issues and a find_name regex utility for identifying AI agent names in issue body JSON. The repo is low-activity (33 stars, last commit 2026-01-22) and the README describes a conceptual architecture that is only partially implemented.

Compared to seeds: no close seed match. The human-as-LLM-proxy concept is unique. Most similar to taskmaster-ai (GitHub Issues as task coordination layer) but inverts the agent/human roles — here the human authors tickets and the AI waits for human approval at every step rather than agents autonomously executing.

ai-ticket — Overview

Origin

Created by jmikedupont2 (Michael DuPont). 33 GitHub stars, MIT licensed, Python, last commit 2026-01-22.

Philosophy

"The AI and Human powered ticket system to manage your AI based code generation with tickets."

"Welcome to our innovative user-driven workflow, designed to empower you to interact with our system like a large language model."

The framework inverts the typical agent-as-executor model: the human takes on the role of an LLM proxy, generating structured queries (tickets) at each step, while AI systems (AutoGPT) wait for the human's ticket responses before proceeding.

Conceptual Architecture

1. User as Language Model: Human generates queries/responses as if they were an LLM
2. Proxy Server: Creates a GitHub Issue ticket for each interaction step
3. AutoGPT Integration: AutoGPT's "Request Assistance" action posts the issue URL to its context
4. Human-Controlled Process: Human writes the response to the ticket; AutoGPT polls for updates
5. Ticket-Based Continuity: Each ticket records one step, ensuring documentation and continuity

Context

This appears to be an experimental research project from early AI agent development (circa 2024). The AutoGPT integration suggests it predates modern Claude Code/Codex-based workflows.

ai-ticket — Architecture

Distribution

Python package with Docker support:

# Docker
docker-compose up -d

# Development
pip install -e .

Required Runtime

• Python 3.x
• Docker (docker-compose.yml present)
• GitHub account + PAT (GITHUB_PAT env var)
• GITHUB_REPO env var

Directory Tree

src/
└── ai_ticket/
    ├── __init__.py       # find_name() regex utility
    ├── backends/
    │   └── pygithub.py   # GitHub Issues integration
    └── events/           # Event handling (minimal)
pyproject.toml
docker-compose.yml
docker-compose-run.yml
.env.example

Config Files

• .env with GITHUB_PAT and GITHUB_REPO

Target AI Tools

AutoGPT (primary integration target per README)

ai-ticket — Components

Python Modules

ai_ticket/__init__.py

find_name(text) — regex utility that extracts AI agent name from a message body matching the pattern:

pattern = r'(```)?\s*{\s*"messages"\s*:\s*\[\s*\{\s*"role"\s*:\s*"system"\s*,\s*\"content"\s*:\s*"You\s+are\s+(?P<name>[^,]+),.*'

Used to identify which AI agent name is referenced in an issue body.

ai_ticket/backends/pygithub.py

GitHub Issues backend using PyGithub library:

No CLI, No Skills, No MCP, No Hooks

ai-ticket is a library/backend, not a standalone agent framework. It provides GitHub Issue integration primitives for higher-level agent systems.

ai-ticket — Uniqueness & Positioning

Differs from Seeds

ai-ticket doesn't match any seed framework closely. It's an experimental concept from early AutoGPT-era AI development (circa 2024) where the human acts as the LLM proxy and AI waits for human ticket responses. The closest seeds are taskmaster-ai (GitHub-based task coordination) and claude-conductor (markdown scaffolding), but neither inverts roles the way ai-ticket does. This is fundamentally a human-in-the-loop at-every-step system, not an AI-autonomous workflow.

Distinctive Positioning

• Human-as-LLM-proxy concept: user authors tickets as if they were an LLM; AI waits
• Step-by-step ticket documentation creates an unusually detailed audit trail
• Designed for AutoGPT era (2024), not modern Claude Code/Codex workflows

Observable Failure Modes

• Incomplete implementation: README describes architecture not fully built
• AutoGPT dependency is outdated relative to modern agent frameworks
• No path for modern AI coding tools (Claude Code, Codex, Cursor)
• Low activity: likely abandoned or superseded by more complete tools