ws-mcp
MCP server that gives LLMs full visibility into your ws-cli workspace tree, enabling queries about workspaces, git status, tasks, and saved browser tabs.
README
ws-mcp
An MCP server that gives LLMs full visibility into your ws-cli workspace tree.
Ask your AI assistant things like:
- "What projects have I abandoned? Summarize where I left off on each one."
- "Which workspaces have uncommitted git changes?"
- "What tasks are open across all my ws/ projects?"*
- "Find everything I was working on related to authentication."
The server traverses your workspace tree, reads metadata, checks git status, parses saved browser tabs, reads beads task databases, and exposes it all through MCP tools — no state of its own.
Prerequisites
- Node.js >= 18
- ws-cli installed and configured (provides the workspace tree that this server reads)
Installation
git clone https://github.com/camggould/ws-mcp.git
cd ws-mcp
npm install
Registering with an MCP Client
Claude Code (per-project)
Add a .mcp.json file to any project directory:
{
"mcpServers": {
"ws-mcp": {
"command": "node",
"args": ["/absolute/path/to/ws-mcp/src/server.js"]
}
}
}
Claude Code (global)
claude mcp add --global ws-mcp node /absolute/path/to/ws-mcp/src/server.js
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"ws-mcp": {
"command": "node",
"args": ["/absolute/path/to/ws-mcp/src/server.js"]
}
}
}
Any MCP-compatible client
ws-mcp uses stdio transport — pipe stdin/stdout to node src/server.js.
Tools
list_workspaces
List all workspaces with metadata. Filter by status, staleness, or tags.
| Parameter | Type | Description |
|---|---|---|
status |
string |
Filter by status: active, paused, archived, abandoned |
stale_days |
number |
Only show workspaces not opened in this many days |
tags |
string |
Comma-separated tags to filter by (matches any) |
list_workspaces({})
list_workspaces({ status: "active" })
list_workspaces({ stale_days: 30, tags: "coding,research" })
Returns: Array of { name, status, last_opened, created, tags, path }.
get_workspace
Deep-dive into a single workspace. Returns everything the server knows.
| Parameter | Type | Description |
|---|---|---|
name |
string |
Workspace name (e.g. "my-project" or "parent/child") |
get_workspace({ name: "ws/cli" })
Returns:
- meta — status, tags, created/last opened dates
- git — branch, dirty/clean, ahead/behind remote, modified/untracked/staged file counts
- tabs — saved browser tabs from last session (count + URLs)
- tab_history — session count, first/last session timestamps, unique URL count across all sessions
- beads — task tracker summary:
total_issues,by_statuscounts,open_by_prioritycounts (ornullif no beads database) - days_since_opened / stale — staleness indicator (stale = 30+ days)
- notes — full
workspace.mdcontent
get_workspace_tree
Full parent-child hierarchy as a nested JSON tree. No parameters.
get_workspace_tree({})
Returns: Nested tree with { name, children[], meta?, path? } at each node.
find_stale_workspaces
Identify workspaces you may have forgotten about, sorted by most stale first.
| Parameter | Type | Default | Description |
|---|---|---|---|
days |
number |
14 |
Days without activity to consider stale |
find_stale_workspaces({ days: 7 })
Returns: Array of { name, status, last_opened, days_since, tags }.
search_workspaces
Full-text search across workspace names, tags, workspace.md notes, and saved tab URLs.
| Parameter | Type | Description |
|---|---|---|
query |
string |
Search query (case-insensitive substring match) |
search_workspaces({ query: "authentication" })
Returns: Array of { workspace, status, matches[] } where each match includes the field name and matching value/context.
summarize_all
High-level dashboard across all workspaces. No parameters.
summarize_all({})
Returns:
- total_workspaces — count
- by_status — breakdown (
{ active: 3, paused: 1, ... }) - recently_active — workspaces opened in the last 7 days
- stale_30_plus_days — workspaces untouched for 30+ days
list_beads
List individual tasks from a workspace's beads issue tracker. Supports filtering and returns tasks sorted by priority then creation date.
| Parameter | Type | Default | Description |
|---|---|---|---|
workspace |
string |
required | Workspace name (e.g. "betterlife" or "ws/mcp") |
status |
string |
Filter: open, in_progress, blocked, closed |
|
priority |
number |
Filter: 0=critical, 1=high, 2=normal, 3=low, 4=trivial |
|
type |
string |
Filter: bug, feature, task, epic, chore |
|
label |
string |
Filter by label (exact match) | |
limit |
number |
50 |
Max issues to return (1–200) |
list_beads({ workspace: "betterlife" })
list_beads({ workspace: "betterlife", status: "open", priority: 1 })
Returns: { workspace, count, issues[] } where each issue includes id, title, description, status, priority, priority_label, type, assignee, labels[], created_at, updated_at.
get_beads_across_workspaces
Aggregate task counts across multiple workspaces. Scope to a parent prefix or query everything.
| Parameter | Type | Description |
|---|---|---|
parent |
string |
Only include workspaces under this prefix (e.g. "ws" for ws/* workspaces). Omit for all. |
status |
string |
Only show workspaces that have at least one issue with this status |
get_beads_across_workspaces({})
get_beads_across_workspaces({ parent: "ws" })
get_beads_across_workspaces({ status: "open" })
Returns:
- aggregate —
total_open,workspaces_with_beads,by_statuscounts,open_by_prioritycounts - workspaces[] — per-workspace breakdown with
by_status,open_by_priority,open_total(sorted by most open issues first)
How It Works
┌──────────────┐ stdio ┌──────────┐ reads ┌──────────────────┐
│ MCP Client │ ◄────────────► │ ws-mcp │ ──────────────► │ ~/Workspaces/ │
│ (Claude, │ JSON-RPC │ server │ │ .workspace.yaml │
│ Cursor, │ │ │ │ tabs.json │
│ etc.) │ │ │ │ tabs-history.jsonl│
└──────────────┘ └──────────┘ │ workspace.md │
│ │ .beads/beads.db │
│ .git/ │
▼ └──────────────────┘
~/.config/ws/
config.json
- Reads
~/.config/ws/config.jsonto find the workspaces root (defaults to~/Workspaces) - Recursively walks the directory tree, following symlinks, looking for
.workspace.yamlmarker files - Parses workspace metadata (YAML), saved tabs (
tabs.json), tab history (tabs-history.jsonl), and notes (workspace.md) - Reads beads SQLite databases (
.beads/beads.db) in read-only mode for task counts and issue listings - Runs
git statuson workspaces that are git repos (via simple-git) - Exposes everything through MCP tools over stdio transport
The server is stateless — it reads directly from the filesystem that ws-cli manages. No database, no cache, no background processes.
Development
# Run with auto-reload on file changes
npm run dev
# Run directly
npm start
Dependencies
| Package | Purpose |
|---|---|
| @modelcontextprotocol/sdk | MCP server framework |
| better-sqlite3 | Read-only access to beads task databases |
| simple-git | Git status checks |
| js-yaml | Workspace metadata parsing |
| fast-glob | File pattern matching |
| zod | Tool parameter validation |
Related
- ws-cli — the CLI tool that creates and manages the workspace tree this server reads
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.