LibrAgent User Documentation
A local agent operating environment. Select your model, install tools with one click, and choose a coordination pattern.You do not need a vendor harness or manual JSON configuration. The software saves completed work as files on your machine.
Language: 한국어 · English · Download Latest Release (Windows / macOS / Linux) · GitHub Repository
Why LibrAgent?
- Direct desktop UI without manual setup: Install one-click MCP presets in Extensions or import configurations from Cursor, VS Code, Claude Code, and Windsurf.
- Built-in multi-agent orchestration: Run standard coordination patterns (
pipeline,hub-spoke,divide-conquer,consensus-delegation,teamwork,schedule) as bundled skills without assembling framework code. - Model and stack freedom (Local-First, MIT): Connect OpenAI, Anthropic, Gemini, Groq, or local Ollama models while keeping your workspace, shell, and browser sessions on your machine.
Get started
- 5-minute tutorial — Install, AI & Models, Chat, App Wizard
- First agent chat
- Connecting models
Guides
- Assistants — Custom agents (General / Tools / Skills)
- Playbooks — Re-run successful workflows
- Solution Recipes — One-click workflow bundles
- Scheduled Tasks · Automation Overview · Session loops (loop) · Continue when done (call-me-back)
- Sessions · Session Export
- Skills · Sub-agents
- Extensions (MCP Tools) · Custom MCP
- Browser automation (built-in browser · your Chrome extension) · Built-in tools
- Navigation Guide · Troubleshooting
FAQ · scenarios
Onboarding in one line
- Settings → AI & Models → Provider API Keys → Save Changes
- Start a session from an assistant card in Chat
- MCP lives under Extensions
- Repeat work with Playbooks / Scheduled Tasks
App UI language follows Settings → General → Language.
Tips
- Settings tab name: AI & Models. MCP: sidebar Extensions (not a Settings tab).
- Contributors: getting-started-dev.md
- Questions: Discussions