Skip to main content

FAQ

Does DesireCore require internet access?​

DesireCore uses a local-first architecture — all core data and Companion repositories are stored locally. However, AI inference requires cloud model API calls, so an internet connection is needed during use.

Which operating systems are supported?​

macOS, Windows, and Linux are supported, along with a Web version.

Is my data safe?​

All Companion data (memory, configuration, skills) is stored on your local filesystem and is never uploaded to third-party servers. AI conversation content is only sent to your configured model API provider.

How do I add new tools to a Companion?​

Integrate external tools via the MCP (Model Context Protocol). Configure them in the project's .mcp.json file.

What are Receipts for?​

Receipts record the complete evidence chain for every Companion execution. You can:

  • See exactly what the Companion did
  • Understand the reasoning behind each decision
  • Roll back to a previous state if something goes wrong

Which AI models are supported?​

DesireCore supports a dual-runtime architecture:

  • Claude Agent SDK: Uses Anthropic Claude model family
  • pi-agent-core: Supports multiple model providers (OpenAI, Google, DeepSeek, MiniMax, etc.)

You can switch runtimes and models in the Agent configuration.

When will the Workflow system be available?​

The workflow system is not included in the MVP phase. The current architecture reserves workflow directories and interfaces — full functionality will ship in a future release.

Having issues?​