wormhole

Wormhole acts as a universal shared memory bridge for AI coding agents, including Claude Code, GitHub Copilot, and Cursor. In modern development workflows, developers often switch between different AI assistants or run multiple sub-agents in parallel, leading to fragmented context and lost information. Wormhole solves this by providing an MCP (Model Context Protocol) server that logs every significant decision, file modification, command execution, and test result into a centralized, persistent repository. This ensures that when you switch tools or agents, your new assistant has immediate access to the full history, established conventions, and recent architectural decisions.

• Centralized Logging: Provides a unified log tool to record file_edit, cmd_run, decision, test_result, todos, plan_output, and feedback.

• Session Management: Organizes work into named, isolatable sessions with support for summary generation and activity tracking.

• Conflict Detection: Actively monitors project paths to alert agents when they are attempting to modify the same files concurrently.

• Knowledge Capture: Features a save_knowledge and search_project_knowledge system to store and retrieve pitfalls, coding conventions, and project-specific constraints.

• Visualization: Includes a web UI via npx wormhole ui for timeline analysis, dashboard stats, and insight filtering.

• Token-Efficient Operations: Designed for AI consumption, utilizing delta-based queries and relevance filtering to minimize context window impact.

• Always use absolute paths for the project_path parameter to avoid scope ambiguity, as the system relies on exact path matching for file integrity.

• Follow the recommended workflow: start_session, pull context with search_project_knowledge, check for conflicts, log actions, save learnings, and conclude with end_session.

• The system performs automated stale event rejection, filtering out edits to files that no longer exist, ensuring the context remains high-fidelity.

• Perfect for CI/CD pipelines, multi-agent refactoring tasks, and long-term project maintenance where history is critical for preventing regressions.

• Easily integrates via MCP configurations in standard editors; use npx wormhole-mcp for instant deployment without complex infrastructure setup.