ling-stack-org/ling-stack

What it solves

Ling Stack addresses the fragmentation of AI-assisted work. When using multiple AI models or agents for a single project, conversations, tool outputs, and artifacts often end up scattered across different browser tabs or applications. This project centralizes these interactions into a single local workspace, keeping the context, history, and results together.

How it works

It operates as a local daemon that manages your AI connections, workspace state, and tool permissions. You connect your preferred AI providers (such as OpenAI, Anthropic, or DeepSeek-compatible endpoints) via a Model Workbench. The system allows you to assign specific models and tools to "teammates" within a project channel, where you can review their activity and approve tool usage before results are finalized. All conversation history and workspace data are stored locally on your machine.

Who it’s for

It is designed for developers and power users who work with multiple AI models and need a unified interface to manage complex tasks, tool-use permissions, and project artifacts without losing track of their workflow.

Highlights

• Unified Workspace: Keeps messages, model selections, tool activity, and generated artifacts in one project-specific channel.
• Provider Agnostic: Supports a wide range of providers, including OpenAI, Anthropic, Volcengine Ark, and other OpenAI-compatible relays.
• Local-First Privacy: Stores workspace state, conversation history, and run evidence locally in a SQLite database.
• Explicit Control: Requires explicit user approval for tool access and external mutations, ensuring you maintain oversight of AI actions.
• Extensible Architecture: Includes support for MCP (Model Context Protocol) and a plugin manager for assigning and activating tools.