FlowState
Provide a persistent, local-first shared memory layer across multiple AI development clients and assistants.
FlowState is a local-first Model Context Protocol (MCP) server designed to synchronize project context across multiple AI assistants and environments, such as Claude, ChatGPT, Cursor, and Grok. Storing all context in a local SQLite database on your machine, it tracks tasks, structural decisions, code changes, and lessons learned without transmitting private project data to third-party clouds. The tool manages multi-agent concurrency by handling task claiming, identifying conflicting decisions, and maintaining a clear audit trail of which model or app made specific changes.
主要功能
01Detailed provenance tracking recording the application, AI model, and workspace for every change
02Local-first SQLite storage keeping all project records, todos, and decisions on your machine
03Automated conflict detection and interactive disagreement resolution cards for clashing decisions
0422 GitHub stars
05Multi-agent task claiming to prevent multiple assistants from duplicating efforts
06Cross-client shared memory across Claude, Cursor, ChatGPT, Grok, and other MCP clients
使用案例
01Coordinating multi-agent development workflows to prevent race conditions on shared tasks
02Maintaining context and continuous project state when switching between different coding assistants like Cursor and Claude
03Auditing architecture decisions and implementation changes made across diverse AI tools