LingYun

Agentic AI assistant for VS Code. LingYun can plan, use tools (files/shell/language features), and execute multi-step tasks in your workspace.

Platform support

LingYun is currently supported on macOS and Linux. Windows is not supported yet because the built-in tools rely heavily on a bash shell environment.

Quickstart

1. Open the LingYun view (Activity Bar → LingYun → Chat).
2. Pick a model from the model dropdown.
3. Describe a goal, for example:
   • “Find where this function is defined and explain it.”
   • “Refactor this file and run tests.”
   • “Search the repo for X and update the docs.”

By default LingYun starts in Plan mode:

• It drafts a short plan first.
• Click Execute to switch to Build mode and let it run.
• Use Stop anytime to cancel.

Plan vs Build

• Plan (read-only): can inspect the workspace (list/glob/grep/read/lsp) and propose steps.
• Build: can edit/write files and run commands. Some tools require approval.

You can toggle modes from the header, and Execute in Plan switches to Build automatically.

Tools & approvals

LingYun uses tools to do real work:

• Workspace discovery: list, glob, grep, read
• Editing: edit, write
• Commands: bash
• Code intelligence: lsp (definition/references/symbols via VS Code)
• Subagents: task (spawn a specialized subagent like explore or general)

When a tool needs approval you’ll see Allow/Deny (and Allow all for the current run). You can also enable:

• lingyun.autoApprove (not recommended)

Subagents (optional)

The agent can delegate work to a separate “subagent” via the task tool. This can help reduce main-session context bloat by running exploration in a separate session and only returning a short summary.

Auto-run the explore subagent before each user turn (runtime-driven prepass):

{
  "lingyun.subagents.explorePrepass.enabled": true,
  "lingyun.subagents.explorePrepass.maxChars": 8000,
  "lingyun.subagents.model": ""
}

Skills (SKILL.md)

LingYun supports reusable “skills” (task playbooks) stored as SKILL.md files with YAML frontmatter (name, description).

• Workspace skills: .lingyun/skills/**/SKILL.md (also scans .claude/skills by default)
• Global skills: ~/.config/lingyun/skills/**/SKILL.md and ~/.claude/skills/**/SKILL.md (only when lingyun.security.allowExternalPaths=true)
• Auto-apply: mention $skill-name in your message to auto-inject that skill’s instructions for the current turn
• Tool: use skill with no args to list, or skill { "name": "..." } to load into the current conversation

Configure discovery with:

{
  "lingyun.skills.enabled": true,
  "lingyun.skills.paths": [".lingyun/skills", "~/.config/lingyun/skills"],
  "lingyun.skills.maxPromptSkills": 50,
  "lingyun.skills.maxInjectSkills": 5,
  "lingyun.skills.maxInjectChars": 20000
}

Sessions

Each chat runs in a session. Use the Session dropdown to switch sessions and + to start a new one.

Optional: persist sessions to disk so they restore after restarting VS Code:

{
  "lingyun.sessions.persist": true
}

Clear persisted sessions with LingYun: Clear Saved Sessions.

Use a local OpenAI-compatible server (optional)

LingYun can connect to local servers that implement the OpenAI API. The server must return models from GET /v1/models.

In VS Code settings:

{
  "lingyun.llmProvider": "openaiCompatible",
  "lingyun.openaiCompatible.baseURL": "http://localhost:8080/v1",
  "lingyun.openaiCompatible.defaultModelId": "your-model-id",
  "lingyun.openaiCompatible.maxTokens": 32000
}

If your server requires an API key, set it in your environment (default env var: OPENAI_API_KEY). You can override the env var name with lingyun.openaiCompatible.apiKeyEnv.

Commands

• LingYun: Start Task (Ctrl+Shift+. / Cmd+Shift+.)
• LingYun: Abort
• LingYun: Clear History
• LingYun: Show Logs

Troubleshooting

• Model shows “Loading…” forever: make sure lingyun.openaiCompatible.baseURL includes /v1 and the server responds to GET /v1/models.
• Responses cut off: increase lingyun.openaiCompatible.maxTokens.
• See what’s happening: run LingYun: Show Logs and enable lingyun.debug.llm.
• Debug plugins: enable lingyun.debug.plugins (workspace plugins live under .lingyun/plugin/).

Advanced

• Custom workspace tools: .vscode/agent-tools.json (schema: schemas/agent-tools.schema.json)
• Workspace plugins/hooks: .lingyun/plugin/*.cjs (see docs/PLUGINS.md)
• Skills: docs/SKILLS.md
• Implementation details and dev notes: AGENTS.md and docs/

Credits

LingYun is inspired by OpenCode. Thanks to the OpenCode open-source project and contributors for making this possible.

License

Apache License 2.0