Automation & Local Intelligence

Rule Engine Automation

Create powerful workflows with the built-in rule engine.

Rule-based triggers: Evaluate every new clip against saved rules.
Flexible conditions: Match on content text, detected formats, tags, or clip types.
Multi-action pipelines: Auto-tag, favorite, pin, copy values, or run multiple actions in sequence.
Local execution: Rules run on-device with no network calls.
Debug logs: Console output shows rule evaluations for easy tuning.

Interact with your data using the on-device LLM.

Local chat experience: Conversational UI powered by the on-device LLM; no cloud API required.
Threaded history: Saved conversations with editable titles and per-thread context.
Clipboard aware: Send clips or summaries into chat for quick reasoning.
Status-aware prompts: Inline status for loading models, disabled chat, or missing model selection.
Safe controls: Disable send while the model is responding; clear and delete threads inline.
Privacy: All processing happens locally on your machine.

ClipKunda includes a built-in Model Context Protocol (MCP) HTTP server to expose its capabilities to external AI clients.

Built-in HTTP server: Optional MCP HTTP server you can start from Settings.
Client templates: One-click config exports for Claude Desktop, Codex Agent, and generic .mcp.json clients.
Tool registry: Exposes ClipKunda tools and resources (clipboard history) to external AI clients.
Streamable HTTP: Uses updated MCP spec (2025-03-26) with streaming support.
Local control: Start/stop server from the app; logs port and status in the console.

Manage your AI models and resources.

Local LLM toggle: Turn MLX-powered metadata on/off.
Model picker: Choose from built-in defaults:
- Llama 3 8B Instruct
- Mistral 7B Instruct v0.3
- Qwen 2.5 7B Instruct
Download manager: Fetch models from Hugging Face directly within the app.
Storage: Models are stored at ~/Library/Application Support/ClipKunda/Models/.