Ling Stack
Bring AI teammates into one local project workspace.
Choose their models. Keep conversations, tool approvals, and results together.
English · 简体中文 · 日本語 · 한국어 · Deutsch · Français · Español · Português (Brasil)
Try a project review · Getting started · Troubleshooting · Architecture · Contributing · Changelog
Note
Source-first alpha. Start with the Web UI, local daemon, and CLI from source. Desktop development is available separately; stable signed installers and a supported production release are not available yet.
Start With One Shared Task
When work involves several models, the task, their responses, tool requests, and results can end up in separate windows. Ling Stack brings that work into a project channel: give teammates model and tool bindings, review their activity, and keep the outcome with the conversation.
Try a bounded project document review with two teammates and a public sample brief. Begin with discussion; add tools and permissions when a later task needs them. Your own provider configuration and API key are required, and provider charges apply.
English interface in a fresh local workspace before configuring a model. This capture shows the workspace layout, not a completed task. Follow the review guide above to try your own first collaboration.
Who Is It For?
Ling Stack is for developers and people who work with several AI models or teammates and want to keep their work in one place. It connects OpenAI, Anthropic, Volcengine Ark, DeepSeek-compatible endpoints, MiniMax, GLM, and OpenAI-compatible relays through a shared Model Workbench.
Workspace state, conversation history, and run evidence are stored locally. Requests and their context go to the model or tool providers you choose. The interface offers 55 language and region options; interface language and AI response language can be configured separately. README translations cover the eight languages linked above, not every interface locale. Other documentation may still be in English or Chinese.
Three Ways To Use It
| Scenario | What you do | What stays together |
|---|---|---|
| Ask a model | Configure a chat model and start a conversation about a document, idea, or project | Messages, model selection, attachments, and results |
| Work with AI teammates | Give teammates model and tool bindings, then collaborate in a project channel | Participants, tool activity, approvals, artifacts, and the final report |
| Create media | Configure an image or video capability and use Media Production | Projects, scenes, generation jobs, and outputs |
Begin with a simple conversation, then add teammates or tools when the work needs them. Skills provide instructions; tool access and permission remain explicit. Coding-agent work uses the Pi runtime adapter, and available execution tools depend on platform setup and workspace policy.
Quick Start From Source
Requirements: Git, Node.js 22.22.1 or newer, and Corepack with pnpm 10.
The repository pins pnpm 10.0.0; Node 22.22.1 is the CI baseline.
1. Install and build the local Web stack
git clone https://github.com/ling-stack-org/ling-stack.git cd ling-stack corepack enable corepack pnpm install --frozen-lockfile corepack pnpm --workspace-concurrency=1 --filter '@ling-stack/cli...' -r build
The filtered build includes the CLI, daemon, Web app, and their workspace dependencies. Desktop and Mobile builds have separate setup requirements.
2. Start the daemon
Keep this command running in the first terminal, from the repository root:
corepack pnpm ling daemon start --foreground
The default daemon address is http://127.0.0.1:17891. Keep it on loopback.
3. Open the Web UI
In a second terminal, from the same repository root:
corepack pnpm ling ui --foreground --open
Open http://127.0.0.1:5173 if the browser does not open automatically. In another
terminal, curl http://127.0.0.1:17891/health should return JSON with "ok": true.
4. Connect a model and try a conversation
Open Settings → Model Workbench, add a chat-capable model with your own API key, and run Check. The first verified chat model becomes the workspace default. Start a new conversation and ask a small question, then inspect the response and its details.
After configuring and checking the model, run:
corepack pnpm ling doctor
Doctor reports model readiness as well as local service health. A fresh workspace with no chat model reports a provider failure even when the daemon and Web UI are running.
Follow Getting started for the first model, conversation, and project channel. See Troubleshooting for setup errors.
Support And Limits
| Surface or integration | Current scope |
|---|---|
| Web, daemon, and CLI | Source alpha; the default getting-started path |
| Electron Desktop | Development shell; platform setup and clean-machine acceptance remain separate gates |
| Chrome extension and Browser Host | Local development integration; Chrome Web Store publication pending |
| MCP and Plugin Manager | Explicit tool assignment, activation, and approval boundaries |
| Office and Computer Use native features | Platform-dependent development; consult the maintained implementation and release gates |
| Image and video generation | Provider- and capability-dependent; selected live adapters still have open gates |
| Mobile | Experimental work outside the current product acceptance scope |
Platform execution, accessibility, signing, and installer acceptance remain open for selected native features. The Implementation Status index records what is implemented and what still needs acceptance. A source build or focused test result does not establish installer readiness.
For questions and reproducible bugs, read Support and use the GitHub Issue forms. Maintenance is best effort. Report vulnerabilities through Security.
Desktop development
Desktop has additional requirements: Xcode Command Line Tools with Swift and codesign
on macOS, or the .NET 8 SDK on Windows for the Office driver. The current Desktop
distribution targets are macOS arm64 and Windows x64. Electron also needs its platform
binary; see Desktop troubleshooting.
corepack pnpm --filter @ling-stack/desktop dev
For platform tools, build commands, and manual checks, follow the Mac and Windows development-package guide. Read Desktop release and acceptance before packaging for distribution. Public source publication and signed binary releases are separate steps.
Data And Permissions
Default workspace state is under .ling-stack/, including ling-stack.sqlite,
sessions/, artifacts/, and logs/. Provider keys entered in Model Workbench are
saved in local workspace storage. Environment and supported keychain references are
also available; see Configuration and data.
Keep keys, databases, session logs, and private outputs out of Git and public reports. Workspace permissions govern file access, shell commands, external mutations, and paid generation. Review approval requests before accepting them. For storage migration, use the CLI storage flows.
Explore Or Contribute
If this workflow is useful to you, star the repository to follow its progress. Tell us what you tried, where the first task got stuck, and whether the result helped through the Issue forms.
- README translation guidelines: help keep the language versions aligned.
- First project review: a small collaboration trial.
- Community launch kit: introduction drafts and first-user feedback plan.
The Web app (apps/web), Desktop shell (apps/desktop), and CLI (packages/cli) share
the daemon (packages/daemon). Core domain logic, contracts, provider adapters, and
runner integrations live under packages/.
- Contributing: setup, focused checks, and pull request expectations.
- Architecture and Daemon API: system boundaries.
- Design index and Implementation Status: maintained contracts and acceptance gates.
- Provider adapters and Pi integration: model and runtime integration.
- Release policy, Changelog, and Third-party notices: publication and licensing.
Ling Stack is available under the MIT License. Community participation follows the Code of Conduct.