Labsco
MCP SERVER

Gorev

by msenol ↗

Keep a task list that survives between AI sessions — templates for creating work, dependencies and subtasks for structuring it, and a browser view of where everything stands.

Project & Task Management✓ Verified
Summary
A task database that several clients share, with the same tasks visible in the browser and the editor.

The daemon is what distinguishes it: one background process holds the workspace's SQLite database, so Claude, Cursor and the VS Code extension all read the same list rather than each keeping its own. The AI-session tools are the other half — an active task, recent history and a context summary — so a new conversation can pick up where the last one stopped. Note the template rule before you plan a workflow: tasks cannot be created freehand, they come from a template.

What it is

A task manager written in Go that runs as a background daemon with a SQLite database per workspace. AI clients talk to it over MCP; the same data is visible in an embedded web UI and an optional VS Code extension.

What you get
  • Tasks created from templates — since v0.10.0 that is the only route — with readable aliases bug, feature, research, refactor, test and doc instead of template UUIDs
  • Listing and filtering by status, project and due date, full task detail rendered as markdown, editing of title, description and priority, and permanent deletion behind a confirmation
  • Status moved between pending, in progress and completed, one task at a time or many in a single batch update
  • Hierarchy: subtasks created under a parent, a task's parent changed or moved to root level, and the whole tree — parents, children and siblings — displayed at once
  • Dependencies declared between two tasks, so one has to finish before another starts
  • Projects created and listed with their task counts, an active project set so new tasks land there automatically, and per-project task views grouped by status
  • Session context for an AI client: an active task that flips itself to in-progress when set, the recently touched tasks, and a context summary of priorities and blockers
  • Search from three directions — full-text with FTS5 and fuzzy matching, natural language queries like 'high priority' or 'tag:bug', and saved filter profiles you can reload, plus search history
  • File paths attached to a task and watched, so status updates can follow file activity, with the watch list and statistics available
  • Everything exported to JSON or CSV and imported back with conflict resolution, and a summary report of task counts, project statistics and recent activity
  • IDE extension management from inside the server: detect VS Code, Cursor and Windsurf, then install, update, check or remove the extension
Requirements

No account and no key — the data is a SQLite file in your workspace. Run it with npx @mehmetsenol/gorev-mcp-server@latest over stdio; the published npm package is @mehmetsenol/gorev-mcp-server, listed at 0.16.3, and Windows needs Node.js installed and on PATH. The daemon starts itself on the first MCP connection and holds a lock file at ~/.gorev-daemon/.lock, which lets several clients share one instance; the web UI comes up at http://localhost:5082. Each workspace gets its own database at .gorev/gorev.db, and GOREV_DB_PATH or GOREV_DATA_DIR moves it. GOREV_LANG chooses en or tr. The project describes itself as under heavy development, with breaking changes expected. Tool names and descriptions are in Turkish.

Setup effort

One command — npx -y @mehmetsenol/gorev-mcp-server@latest