MCP Server

The Wórkiro MCP (Model Context Protocol) server enables users to interact with their tasks, documents, and projects through AI assistants like Claude, ChatGPT, and Microsoft Copilot. This integration transforms how professionals manage their work, replacing clicks and navigation with natural conversation.

Understanding MCP

What Is MCP?

MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems. Introduced by Anthropic in November 2024 and now governed by the Linux Foundation, MCP has been adopted by all major AI platforms including OpenAI, Google, and Microsoft.

The simplest way to think about MCP: It's like USB-C for AI assistants.

Just as USB-C provides a universal way to connect devices, phone to charger, laptop to monitor, camera to computer, MCP provides a universal way for AI assistants to connect to business applications.

Why Does This Matter?

Before MCP, every AI integration required custom development. Connecting Claude to Google Drive was different from connecting it to Slack, which was different from connecting it to Salesforce. This created what developers call an "N×M problem", each AI application needed separate integration work for each data source.

MCP solves this by establishing a common protocol. Build one MCP server for your application, and it works with every AI client that supports MCP.

Who Supports MCP?

MCP is supported by:

• Claude (Anthropic) — via Claude.ai and Claude Desktop

• ChatGPT (OpenAI) — via desktop and mobile apps

• Gemini (Google) — via Google's AI products for developers

• Microsoft Copilot — across Microsoft 365

• Development tools — Cursor, Replit, VS Code, and others

This broad adoption means Wórkiro users can choose their preferred AI assistant while maintaining the same seamless access to their work.

The Wórkiro MCP Server

What We've Built

The Wórkiro MCP Server is a secure connection layer that exposes Wórkiro's core functionality to any MCP-compatible AI assistant. Users authenticate once, and their AI assistant gains the ability to read and interact with their Wórkiro account via MCP Tools.

What Are MCP Tools?

In MCP terminology, Tools are the specific actions an AI assistant can take when connected to an application, essentially the building blocks of everything the AI can read, retrieve, and do on your behalf. When you connect Wórkiro to your AI assistant, it receives a menu of available Tools and decides which to use based on what you ask. Ask "What do I need to do today?" and the AI calls the task retrieval tool. Say "Create a task for Sarah" and it calls the task creation tool. You never see this happening, you just get a response. The Wórkiro MCP Server currently exposes 17 tools across four functional areas:

Your Tasks and Threads

• Read actions

  • get_my_tasks — Retrieves all tasks currently assigned to you, including optional detail on participants, comments, attachments, and related connections or projects

  • get_threads — Retrieves one or more specific threads (tasks, notes, or signature requests) by ID, with participants and related entities

  • get_thread_comments — Retrieves the full comment history on any thread, useful for understanding context and conversation before acting

• Write actions

  • task_complete — Marks a task as complete

Connections - Your Clients and Contacts

• Read actions

  • find_or_get_connection — Searches for or retrieves a specific connection (person or organisation), and can optionally pull in their related documents, projects, messages, and contacts in a single call — the engine behind client briefings

• Write actions

  • create_person — Creates a new individual contact in Wórkiro, with optional relationship type assignment

  • create_organisation — Creates a new organisation connection, with optional relationship classification

  • create_relationship / update_relationship — Define and manage connection relationship types such as client, supplier, partner.

Workspaces and Projects

• Read actions

  • get_workspaces — Lists all workspaces in your Wórkiro account

  • get_workspace — Retrieves a specific workspace and its configuration

  • get_workspace_tags — Retrieves the classification tags available within a workspace

  • get_workspace_unit — Retrieves a specific project or unit, including its connected documents, messages, and related connections

  • get_workspace_units — Retrieves multiple projects or units in a single call

• Write actions

  • create_project — Creates a new project within a workspace, linked to relevant connections

Tasks — Creating and Supporting

• Read actions

  • get_relationships — Retrieves the relationship types available in your account (such as Client, Supplier, Partner), used when creating or classifying connections

  • get_profile — Retrieves the current user's Wórkiro profile, allowing the AI to personalise responses and correctly attribute actions

• Write actions

  • create_task — Creates a new task in Wórkiro, with optional assignment, due date, participants, related connections, project links, and tags

  • add_comment — Creates a new comment on a thread in Wórkiro.

Search

• Read actions

  • search — A global search across your Wórkiro account spanning connections, projects, documents, and tasks — the AI's starting point for any open-ended query where it needs to locate the right record before doing something with it

Write actions are disabled for an account by default. Only the account owner/admin has permission to configure write actions via an ‘Interactive Tool' (MCP App) which can be used to update the policy for the team. This tool is available in the MCP server as Configure_tool_access

Planned

• The above tools reflect the V1 of Wórkiro MCP and amongst other things we plan to extend these to include more document actions including.

  • Read the text content of a Wórkiro file

  • Download a file from Wórkiro

• Other document actions (for example uploading documents) are covered by the creation of Wórkiro specific agent skills, leveraging our documented Hyperlink API - See below.

Agent Skills: Turning Tools Into Workflows

If tools are the individual actions an AI can take, Agent Skills are the playbooks that tell it how to combine those actions intelligently to achieve something useful. A Skill packages up a multi-step workflow, the sequence of tools to call, the order to call them in, how to handle what comes back, and how to present the result, so the AI can complete a complex task from a single natural language request, without the user needing to guide it step by step. The same Skills work across different AI agents: a Skill written for Claude works equally in ChatGPT or Copilot, because they all speak the same MCP language underneath.

We are building a library of Wórkiro Agent Skills, with two available for use today - get in touch via help@workiro.com

Getting Started

Requirements

A Wórkiro account (paid plan) and AI assistant subscription.

Available now...

• Claude.ai (Pro, Team, or Enterprise)

Coming soon...

• ChatGPT (Plus or Team)

  • The Wórkiro MCP connection is in development
     

• Microsoft Copilot

  • The Wórkiro Agent is in development

  • The Wórkiro MCP Connector will also be made available via Microsoft

Setup

1. Request MCP server access for your Wórkiro account via help@Wórkiro.com

2. Add Wórkiro MCP server to your AI assistant's configuration

   • Configure Wórkiro MCP Server for Claude 

   • Configure Wórkiro MCP Server for Chat GPT (Soon) 

3. Install Wórkiro specific skills (Contact us)

4. Individuals are authenticated when prompted

5. Start asking questions about your work

FAQ

Q: Can the AI assistant access everything in my Wórkiro account? A: The AI model accesses data according to your existing Wórkiro permissions. It cannot see content you don't have access to, and team access controls are fully respected.

Q: Does this work with the mobile apps? A: MCP support varies by AI provider. Claude and ChatGPT currently support MCP on desktop applications, once connected these are available on their mobile apps.

Q: Is my data sent to the AI provider? A: When you make a request, relevant data is sent to the AI to formulate a response. All major AI providers have enterprise-grade security and data handling policies. Wórkiro's data remains in Wórkiro; only query results are shared with the AI.

Q: Can I use this with multiple AI assistants? A: Yes. You can connect Wórkiro to multiple MCP-compatible AI assistants simultaneously.

Security Model

• OAuth 2.0 Authentication: Users authenticate with their existing Wórkiro credentials

• Permission Inheritance: AI access respects all team access controls from the user’s account.

• Audit Trail: All MCP tool interactions are logged internally against the account and user.

• Data Minimisation: AI receives only the data needed to fulfil each request.