# WCP Cloud

Work continuity for AI agents. Work items, status, decisions, and history
that survive across sessions. Your agent reads and writes to it as you work.

Website: https://workcontextprotocol.io
GitHub: https://github.com/dpaola2/work-context-protocol
MCP Endpoint: https://workcontextprotocol.io/mcp


## Quick Start

### Claude Code

claude mcp add --transport http --scope user wcp https://workcontextprotocol.io/mcp

### Codex CLI

codex mcp add wcp --url https://workcontextprotocol.io/mcp

### Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "wcp": {
      "url": "https://workcontextprotocol.io/mcp"
    }
  }
}

### Other MCP-compatible agents

Point your agent's MCP configuration at: https://workcontextprotocol.io/mcp

After connecting, your agent will prompt you to authenticate via
browser. Once verified, WCP tools are available in every session.


## MCP Tools

WCP exposes 27 tools via the Model Context Protocol:

### Namespace Management

wcp_namespaces              List all namespaces with item/document counts
wcp_create_namespace        Create a new namespace (e.g. PROJ, OPS)
wcp_rename_namespace        Rename a namespace key, updating all callsigns
wcp_update_namespace        Update namespace properties (archive/unarchive, name, description)
wcp_schema                  Discover valid field values (status, priority, type)
wcp_schema_update           Add custom statuses or artifact types per namespace

### Work Items

wcp_list                    List items in a namespace (filter by status, priority, type, project, assignee, parent)
wcp_get                     Get item content by callsign (e.g. PROJ-12). Supports offset/limit for partial body reads
wcp_create                  Create a new work item, returns auto-generated callsign
wcp_update                  Update any field on a work item
wcp_bulk_update             Update multiple work items at once (status, priority, assignee, etc.)
wcp_comment                 Append to an item's activity log

### Artifacts

wcp_attach                  Attach a document (PRD, spec, proposal) to a work item
wcp_get_artifact            Retrieve an attached artifact's content. Supports offset/limit for partial reads
wcp_approve                 Record approval/rejection on an artifact

### Documents (Knowledge Base)

wcp_doc_create              Create a standalone document (architecture decisions, meeting notes, patterns)
wcp_doc_get                 Retrieve a document by NS/slug reference (e.g. PROJ/roadmap). Supports offset/limit for partial reads
wcp_doc_update              Update a document's title or body
wcp_doc_rename              Change a document's slug
wcp_doc_delete              Delete a document by NS/slug reference
wcp_doc_list                List documents in a namespace, optionally filtered by parent item

### Graph Navigation

wcp_links                   Get forward links and backlinks for any item or document
wcp_resolve                 Entity resolution — find items/docs by name, callsign, or title
wcp_mentions                Discover unlinked references to an entity across all content

### Collaboration

wcp_share_namespace         Generate a share link for a namespace you own
wcp_revoke_collaborator     Remove a collaborator's access to a namespace
wcp_inbox                   List activity events from shared namespaces


## Core Concepts

- Namespaces organize work by domain (PROJ, OPS, etc.)
- Work items have callsigns (PROJ-12), structured fields, a markdown body, and an activity log
- Documents store prose knowledge, optionally linked to work items
- Artifacts are files attached to work items (PRDs, specs, proposals)
- Activity logs are append-only comment threads on each item
- Links connect items and documents into a navigable knowledge graph


## Linking

WCP automatically parses wiki-links and callsign references from content
when items, documents, or comments are created or updated. Use wiki-link
syntax to create explicit connections:

  [[PROJ-12]]         Link to a work item by callsign
  [[PROJ/roadmap]]    Link to a document by NS/slug reference
  [[Some Title]]      Link to a document by title (case-insensitive)

Bare callsign references (e.g. PROJ-12 without brackets) are also
detected automatically. Link counts (incoming/outgoing) appear in
wcp_get and wcp_doc_get responses.

Use wcp_links to navigate the graph before reading full content — it
returns connections in ~200 tokens vs ~5-10K from reading bodies.
Use wcp_resolve to find entities by any reasonable name. Use
wcp_mentions to discover where something is referenced but not
formally linked.


## Partial Reads

For large content, use offset and limit to read specific line ranges
instead of loading everything into context:

wcp_get_artifact(id: "PROJ-5", filename: "plan.md", offset: 1, limit: 50)
wcp_doc_get(ref: "PROJ/roadmap", offset: 100, limit: 25)
wcp_get(id: "PROJ-5", offset: 10, limit: 30)

- offset: 1-indexed start line (defaults to 1)
- limit: number of lines to return (defaults to all remaining)
- Partial reads include metadata: totalLines, returnedLines, offset, limit
- Omitting both returns full content (backward compatible)
- wcp_get offset/limit applies to body only; activity log and artifacts always return in full
- Summary responses include line counts (bodyLines, artifact lines, doc lines) so you can check size before reading


## Field Values

status:     backlog, todo, in_progress, in_review, done, cancelled (extensible per namespace)
priority:   urgent, high, medium, low
type:       feature, bug, chore, spike


## When to Use WCP

Use WCP when you need:
- Persistent project context across sessions
- Work tracking that happens as a side effect of doing the work
- Shared context between team members' agents
- A knowledge base your agent can read and write to naturally

Do not use WCP when you need:
- Source code storage (use git)
- File system access (WCP stores work metadata only)
- Real-time collaboration (WCP is async, built for agent workflows)


## Example: Agent Workflow

# Agent discovers available namespaces
wcp_namespaces
# Returns: [{key: "PROJ", name: "My Project", items: 12, documents: 3}]

# Agent lists active work
wcp_list(namespace: "PROJ", status: "in_progress")
# Returns: [{id: "PROJ-7", title: "Add user auth", status: "in_progress"}]

# Agent reads full context before starting work
wcp_get(id: "PROJ-7")
# Returns: frontmatter, body, activity log, artifacts list

# Agent logs progress as it works
wcp_comment(id: "PROJ-7", author: "claude", body: "Implemented JWT middleware, tests passing")

# Agent marks work complete
wcp_update(id: "PROJ-7", status: "done")

# Agent creates a new item for follow-up work
wcp_create(namespace: "PROJ", title: "Add refresh token rotation", parent: "PROJ-7")
# Returns: {id: "PROJ-13"}

# Agent navigates the knowledge graph
wcp_links(id: "PROJ-7")
# Returns: forward links, backlinks, counts — understand connections without reading bodies

# Agent resolves an entity by name
wcp_resolve(name: "auth spec")
# Returns: matching documents across accessible namespaces

# Agent discovers unlinked references
wcp_mentions(id: "PROJ-7")
# Returns: places where PROJ-7 is mentioned but not formally linked


## Data & Privacy

- Work metadata only: titles, status, comments, documents
- No source code access, no filesystem access
- Open source protocol, MIT-licensed
- Migrate between cloud and self-hosted anytime
- Data stored on US infrastructure


## Pricing

Free for individuals. Hard spending caps.


## Built By

Dave Paola — Pittsburgh, PA
https://github.com/dpaola2