Show and Tell: ORCH — terminal-first alternative for teams who prefer code over visual workflow builders

[oxgeneral]

Hey FastGPT community 👋

FastGPT's visual workflow builder is great for teams that prefer a GUI. But for teams who live in the terminal and want to orchestrate AI agents via TypeScript code and CLI commands, I built ORCH — a CLI runtime for managing AI agent teams.

The use case:
Instead of drawing boxes in a UI, you define agent roles, assign tasks, and let the state machine handle coordination:

# Define your team (YAML config)
orch agent add "backend-dev" --adapter claude --role "Senior TypeScript developer"
orch agent add "code-reviewer" --adapter claude --role "Code reviewer"

# Add tasks with dependencies
orch task add "Implement auth endpoints" --assignee backend-dev
orch task add "Review auth implementation" --assignee code-reviewer --depends-on tsk_xyz

# Watch it run
orch tui

What ORCH provides:

• Formal state machine: todo → in_progress → review → done
• Mandatory review gate — no code is marked done without review
• Auto-retry when agents fail (configurable attempts)
• Inter-agent messaging so agents coordinate in real time
• Shared context store for passing structured data
• Full programmatic API via @oxgeneral/orch npm package

For FastGPT users: Do you have use cases where you want to trigger your FastGPT workflows from CLI scripts or CI/CD pipelines? I'm curious how teams integrate visual builders with their existing DevOps workflows.

GitHub: https://github.com/oxgeneral/ORCH
npm: npm install -g @oxgeneral/orch

[Markzhangeth]

Hello, I see you're working on the FastGPT project. We have a multi-model API transfer service compatible with OpenAI formats, supporting GPT, Claude, Gemini, Kimi, and other models, suitable for developers to quickly integrate and reduce the cost of maintaining multiple models. If your project has API call requirements, I can give you some test credits to try it out. One-stop shop for large model APIs: https://xtokens.shop