Summary
An open-source desktop platform for running AI agents locally and connecting them to models, tools, and workflows.
About
Craft Agents combines locally run AI agents, connections to external and local tools, and reusable workflow features in a desktop interface. Its provider flexibility and local execution may suit teams seeking control over model choice and data handling.
Craft Agents is an open-source desktop application that runs on the user's machine, keeping data and provider keys local. It supports Claude, GPT, Gemini, and local models. Agents can connect to MCP servers, APIs, and local files to perform tasks across those resources. The interface renders code diffs, data tables, diagrams, and live previews inline. Users can save results as persistent Pages, create reusable Skills, and schedule Automations.
Key Features
Open-source desktop application that runs locally
Support for Claude, GPT, Gemini, and local models
Agent connections to MCP, APIs, and local files
Inline rendering of code diffs, data tables, diagrams, and live previews
Persistent Pages, reusable Skills, and scheduled Automations
Real-World Use Cases
Developer task orchestration
Connect agents to code-related tools and local files, review code diffs inline, and save repeatable workflows as Skills or scheduled Automations.
Research and analysis workflows
Use agents with APIs and local files to gather or process information, then inspect results in tables, diagrams, and persistent Pages.
Internal workflow automation
Create reusable agent procedures and schedule recurring tasks that operate across connected tools.
Business Opportunities
Developer productivity workflows
Problem: Development teams may need to coordinate AI assistance across code, local files, and connected tools while reviewing generated changes.
Solution: Use locally run agents with tool connections and inline code diffs, then standardize recurring tasks as reusable Skills.
Estimated ROI: Potentially moderate through reduced manual coordination and reuse of common workflows; actual returns depend on adoption and task suitability.
Execution: Identify repeatable development tasks, connect the relevant MCP servers, APIs, or local files, and create Skills for workflows that benefit from consistent execution.
Scheduled operational automation
Problem: Recurring information-processing tasks can require repeated manual work across multiple tools and data sources.
Solution: Configure agents to work with connected APIs and files, then schedule suitable recurring tasks as Automations.
Estimated ROI: Potentially moderate to high for frequent, repeatable tasks, subject to setup effort, reliability, and human review requirements.
Execution: Select stable, well-defined recurring tasks, validate agent outputs, and schedule automations only after confirming the connected tools and data sources behave reliably.