Autonomous AI dev-team bot for automating software development tasks. It enables users to describe features via Telegram, which are then planned, implemented, tested, reviewed, and submitted as pull requests by an AI team powered by Claude Code. This offers engineers and power users a streamlined, AI-driven development workflow with isolated workspaces.

How It Works

The project leverages the Claude Code CLI, driven by a native Go service, to form a 5-role AI dev team: planner, coder, tester, reviewer, and arbiter. This team executes a pipeline from feature specification to an adversarial code review with inline comments, culminating in a pull request. Each Telegram chat session operates within its own isolated workspace (/workspace/chat_<id>), allowing parallel development across different conversations. PR reactions are managed via a polling mechanism that queries the git host, ensuring functionality even with network restrictions.

Quick Start & Requirements

The recommended installation uses Docker. Clone the repository, navigate to adapters/telegram, copy .env.example to .env, and fill in the required variables: TELEGRAM_BOT_TOKEN, TELEGRAM_BOT_USERNAME, ALLOWED_USERS, and CLAUDE_CODE_OAUTH_TOKEN (or ANTHROPIC_API_KEY). Then, run docker compose up -d. The bot requires hosting in an Anthropic-supported region. Git host integration (Gitea, GitHub, GitLab) requires GIT_HOST, GIT_USER, and GIT_TOKEN in .env.

Highlighted Details

• Autonomous 5-role AI dev team (planner, coder, tester, reviewer, arbiter) with a defined pipeline.
• Per-chat isolation ensures distinct, parallel development workspaces.
• PR polling mechanism handles review comments reliably without inbound webhooks.
• Optional voice message transcription and execution via Mistral, OpenAI, or local models.

Maintenance & Community

No specific community channels (e.g., Discord, Slack) or notable contributors are detailed in the provided README.

Licensing & Compatibility

The project is licensed under the MIT license, which generally permits commercial use and integration with closed-source projects.

Limitations & Caveats

The service must be hosted in an Anthropic-supported geographical region. The AI agent is granted shell and edit access within its containerized sandbox, necessitating strict adherence to the ALLOWED_USERS whitelist for security. The Git token used for repository access is shared across a deployment and should be scoped accordingly.