Connecting Claude via MCP

TskVlo includes a built-in MCP (Model Context Protocol) server that lets AI assistants like Claude read and write your project data in real time. This guide walks through the complete setup and shows you how to start managing tasks through natural conversation.

What Is MCP?

MCP is an open standard created by Anthropic for connecting AI language models to external tools and data sources. The AI client (Claude) connects to an MCP server (TskVlo), which exposes a set of named tools. When you ask the AI to do something that involves your tasks, it calls the appropriate tools, and the results flow back into your conversation.

The key benefit over traditional integrations is that the AI has structured, live access to your data — not just text you’ve copied in. It can query current state, make changes, and chain multiple operations together based on a single instruction.

Prerequisites

• A TskVlo account with the Owner or Admin role in the workspace you want to connect — only they can authorize an AI client or create API keys
• An MCP-compatible AI client: Claude (claude.ai, Claude Desktop, Claude Code), Cursor, or any other client that supports remote MCP servers

The MCP Server URL

Whichever way you connect, the TskVlo MCP server lives at:

https://dash.tskvlo.app/api/mcp

You’ll also find it in TskVlo under Workspace settings → API keys, in the MCP Server section.

Two Ways to Connect

Option A: Connect via OAuth (Recommended)

If your client supports OAuth for remote MCP servers — like Claude’s custom connectors — you don’t need to handle any keys.

1. In Claude, open Settings → Connectors and choose Add custom connector
2. Name it TskVlo and enter the URL https://dash.tskvlo.app/api/mcp
3. Click Connect — a TskVlo page opens in your browser. Log in if you aren’t already (two-factor authentication works as usual)
4. On the Authorize MCP access screen, pick the Workspace the connection should use. Only workspaces where you are Owner or Admin are listed
5. Review the requested permissions. To let the AI read but never change anything, tick Grant read-only access
6. Click Authorize — you’re sent back to Claude, and the TskVlo tools are available

Behind the scenes, the client registers itself with TskVlo automatically and signs in with PKCE. Access tokens are short-lived and refreshed in the background, so you won’t be asked to log in again while you keep using the connection.

Other OAuth-capable clients work the same way: add https://dash.tskvlo.app/api/mcp as a remote (HTTP) MCP server, and the client will start the sign-in flow on first use. In Claude Code, for example:

claude mcp add --transport http tskvlo https://dash.tskvlo.app/api/mcp

Then run /mcp inside Claude Code to authenticate.

Ending an OAuth Connection

• Remove or disconnect the connector in your AI client to stop it from using TskVlo.
• A connection stops working as soon as you lose the Owner or Admin role in that workspace, or are removed from it.
• Connections that aren’t used for 30 days expire on their own; you’ll simply be asked to authorize again.

Option B: Connect with a Workspace API Key

For clients without OAuth support, or for your own scripts:

Step 1: Create an API Key

1. Open Workspace settings → API keys (Owners and Admins only)
2. Click Create key and give it a name, e.g. “Cursor”
3. Choose the Access:
   • Read & write — can read and change boards, items, and wiki pages
   • Read-only — can only read; write requests are rejected and MCP write tools are hidden
4. Click Create key and copy the key immediately — it won’t be shown again

API keys are scoped to the workspace they were created in. You can see when each key was last used and revoke it at any time from the same page.

Step 2: Add the Server to Your Client

Clients that take a JSON configuration with a URL and headers — Cursor’s mcp.json, for example — look like this:

{
  "mcpServers": {
    "tskvlo": {
      "url": "https://dash.tskvlo.app/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Replace YOUR_API_KEY_HERE with the key you copied. Instead of the Authorization header you can also send the key as x-api-key: YOUR_API_KEY_HERE.

In Claude Code:

claude mcp add --transport http tskvlo https://dash.tskvlo.app/api/mcp \
  --header "Authorization: Bearer YOUR_API_KEY_HERE"

Restart or reload your client after changing its configuration.

Verify the Connection

Start a new conversation and type:

List the boards in my TskVlo workspace.

If the connection is working, the AI will call the list_boards tool and return your boards. If you see an error, check the troubleshooting section below.

Available Tools

Once connected, the AI has access to around 40 TskVlo tools:

Read-only connections only see the reading tools. Changes made through MCP — new items, updated automation rules, notifications marked read — show up in the open TskVlo app without a reload.

Example Prompts

Here are prompts that work well once connected:

Querying data:

• “What boards are in this workspace?”
• “Show me all items assigned to me that are due this week.”
• “What’s on the ‘Onboarding’ wiki page?”

Creating and updating:

• “Create a new board called ‘Q3 Roadmap’.”
• “Add a bug report: title ‘Login button unresponsive on Safari’, assign it to Emma, set priority to High.”
• “Move the ‘Invoice export’ task from the Support board to the Backlog group of the Engineering board.”

Notifications (OAuth):

• “What are my unread notifications? Group them by board.”
• “Mark all notifications about the Marketing board as read.”

Automations:

• “List the automation rules on the Support board and explain what each one does.”
• “Create a rule on the Sales board: when an item is created in ‘New leads’, assign it to the person who created it.”

More ideas in the AI prompts guide.

Tips for Best Results

Be specific about board and group names. “The Backlog group on the Marketing board” is unambiguous.

Use it for batch work. Where clicking through a UI is tedious — creating ten similar items, bulk-updating statuses, or adding the same field value to a group of tasks — the AI can do it in a single instruction.

Ask for confirmation before destructive operations. Before bulk-deleting or bulk-moving items, ask the AI to list what it’s about to change first.

Start read-only. If you just want answers and reports, connect read-only. You can always authorize a second, read-write connection later.

Chain context. The AI remembers the conversation, so you can say “Now create three sub-tasks under the first item” after listing items, and it will know what “the first item” refers to.

Troubleshooting

The consent screen shows no workspace: You need the Owner or Admin role in at least one workspace to authorize an MCP client. Ask a workspace owner to change your role, or to set up the connection for you.

“Unauthorized” error: The API key may have been revoked, or the OAuth connection expired or lost its permissions. Create a new key or reconnect the connector.

The AI can’t create or change anything: The connection is read-only. Create a Read & write key, or reconnect via OAuth without ticking Grant read-only access.

Notification tools are missing or fail: They need an OAuth connection. API keys belong to no user and have no inbox.

The AI doesn’t seem to use TskVlo tools: Restart your client fully. If you use a JSON config, check that it is valid (no trailing commas, correct brackets).

Changes aren’t reflected: TskVlo data is live — if the AI queries immediately after a change, it should see the updated state. If you’re seeing stale data, ask it to re-fetch.