OpenBMB/PilotDeck

What it solves

PilotDeck addresses the limitations of current AI agents in long-running, multi-project productivity work. It solves the problem of "context pollution" where different projects bleed into each other, the high cost of using flagship models for simple tasks, the lack of transparency in how agents remember information, and the "ask-and-answer" loop that requires constant user presence.

How it works

It operates as an agent operating system based on the concept of a "WorkSpace," which isolates files, memory, and skills for each project. It employs three core mechanisms:

• White-box Memory: Makes memory generation and retrieval visible and editable, allowing users to fix errors or rollback states via a "Dream Mode."
• Smart Routing: Automatically detects task difficulty to route complex requests to flagship models and simple ones to lighter models, reducing token costs.
• Always-on Execution: Allows agents to run in the background to discover tasks and produce deliverables as local files without active user input.

Who it’s for

It is designed for users engaged in general-purpose, multi-task productivity work, such as AI engineers, content creators (e.g., social media operations), and developers who need an agent capable of handling long-horizon tasks across multiple isolated projects.

Highlights

• WorkSpace Isolation: Dedicated file systems and memory stores per project to prevent cross-project interference.
• MCP Native: Full integration with the Model Context Protocol (MCP).
• Traceable Memory: End-to-end visibility into memory extraction and storage with manual editing capabilities.
• Cost Optimization: Significant token savings through automated model routing (e.g., up to 70% on specific workloads).
• Background Autonomy: Ability to proactively monitor and land results as files on disk while the user is away.