Pi-Wm Pi-Wm

Pi-Wm brings conversations, project files, Git changes, terminals, browser previews, and multi-agent collaboration into a single workspace. You can connect your own model services to handle everything from understanding requirements and writing code to running and verifying it within your local project. Local-first does not mean running offline. Workspace operations and session states are managed by local services; when using remote models, requests and related context are still sent to your configured provider. Core Features Complete Work Around Projects - Local Coding & Terminal: Read and edit project files, view Git diffs, run commands in your local shell, and inspect results using the integrated terminal. - Browser Verification: Agents can view pages, interact with controls, take screenshots, and inspect errors via a standalone Playwright browser; the desktop client also features a built-in browser for you to directly view local previews. - Session & Task Management: Save conversation history, search chat messages, and view request durations, failure details, token usage, and context consumption. - Recovery & Notifications: Provides post-interruption resumption, context compression, and desktop task notifications, making it easy to continue unfinished work. From Individual Tasks to Team Collaboration - Subtasks: Delegate independent tasks to sub-agents, and view their complete conversations, execution states, and results. - Agent Teams: A lead organizes continuously running team members, collaborating through shared tasks, dependencies, and member messaging; the workspace displays team progress and acceptance records. - Member Templates: Save and reuse member roles, models, and tool configurations to minimize repetitive setup. - Goals & Scheduled Tasks: Supports goal acceptance, revision rounds, and one-time or fixed-interval scheduled tasks. Tasks will not continue executing once the application is closed. Models, Tools, and Desktop Capabilities | Capability | What You Can Do | | ------------ | ------------------------------------------------------------------------------------------ | | Model Configuration | Configure your own model services, model capabilities, context windows, and thinking options. | | Official Accounts | Local mode provides authorization portals for Claude, ChatGPT, and Grok; available models and quotas depend on account permissions and provider restrictions. | | Images & Video | Configure dedicated media models to generate and view media results within conversations; actual capabilities depend on the connected service. | | MCP | Add, edit, authorize, and connect MCP services in the interface; saving a configuration does not automatically grant trust. | | Skills | Use built-in skills, and install, preview, and manage local skill packages from workspace directories. | | Computer Use | Once explicitly enabled on Windows, operate desktop applications via UI Automation or screenshots and input; requires a Python 3 environment. | | Interface & Themes | Supports English and Chinese interfaces, light and dark color schemes, and theme settings that follow the system. | For capability boundaries and validation scopes, please refer to the dedicated documentation below. Autonomous planning in Agent Teams requires a real model; Demo mode will not simulate model decision-making.