Skip to content

Personal Assistant

Bob Lee edited this page Jul 27, 2026 · 2 revisions

[English] | 简体中文 · Home

Claw is BitFun’s assistant experience for work that benefits from continuity. Instead of starting from a blank generic chat every time, you can create distinct assistants with their own identity, role, context, capabilities, and memory.

More than one assistant

Different responsibilities should not compete for the same personality or context. You might keep:

  • a weekly planning assistant;
  • a product research assistant;
  • a writing and editing partner;
  • a project operations assistant;
  • a personal learning coach;
  • a monitoring or briefing assistant.

Each assistant lives in its own workspace and can be configured independently.

Give it a clear identity

Define how an assistant should think and communicate:

  • name, role, and visible identity;
  • core personality and working principles;
  • information about the user it serves;
  • preferred tone and collaboration style;
  • primary and fast model choices;
  • selected tools, Skills, and MCP connections.

This makes behavior more consistent than repeatedly rebuilding the same context in prompts.

Long-term memory—with controls

BitFun can extract and consolidate useful knowledge from completed local sessions, then make a compact memory summary available to future work. You can control whether memory is enabled, how it is processed, how external context is handled, and when it should be reset.

Memory is meant to preserve durable preferences and context—not to remove your ability to inspect or clear it.

Scheduled work

An assistant can be prepared for recurring or future sessions. This is useful for regular planning, periodic research, reminders that need agent work, or repeated project routines.

Scheduled work should still be scoped thoughtfully: external systems, important changes, and consequential actions may require their own permissions or confirmation.

Session insights

BitFun can analyze recent sessions across a chosen period and turn activity into a personal report. Insights can include:

  • goals and completed work;
  • wins and recurring friction;
  • working patterns and active times;
  • commonly used tools and project areas;
  • suggestions for improving the next cycle.

This is designed as a reflective layer over your own sessions, not a public productivity score.

Voice input

Speak naturally when typing is inconvenient. BitFun supports configurable local and cloud speech recognition, microphone diagnostics, and multiple recognition choices. Local models can keep transcription on the device; cloud options can be useful when you prefer a hosted service.

An assistant you can see

For people who enjoy a more expressive desktop, BitFun can show an agent companion inside chat or as a floating always-on-top character. Choose a preset or import a compatible companion pack. It is optional—the underlying assistant works the same without it.

Make the assistant as capable—or as focused—as needed

An assistant does not need every available capability. A focused setup is often better:

  • give a research assistant web access and citation-oriented Skills;
  • give a writing assistant document and presentation abilities;
  • give a project assistant selected workspace and scheduling tools;
  • keep a sensitive assistant read-only;
  • use project-level permissions for shared environments.

The result is not one universal bot. It is a set of assistants shaped around the work and trust level you choose.

Next: customize capabilities, review security and privacy, or get started.

Clone this wiki locally