MCP Tools

MCP (Model Context Protocol) is ThinkWork’s connector path for external tool systems. It lives under Connectors, not as a separate platform pillar.

Use MCP tools when agents need to call remote capabilities during work, such as CRM actions, ticket updates, internal API operations, search, or database queries exposed by an MCP server.

Note

For the broader connector model, start with Connectors Overview. For app and channel integrations like Slack, GitHub, and Google Workspace, see Integrations.

How MCP tools fit

ThinkWork uses two broad connector paths:

1. Integrations bring external messages and events into threads
2. MCP tools give agents external capabilities they can call during a turn

Both still operate inside the same ThinkWork model:

• threads remain the unit of work
• agents decide what to do
• controls govern access and usage
• audit trails capture tool activity

How it works

Admin registers MCP server
    → Assigns it to agent template or agent
    → Agent invocation resolves auth
    → Agent container connects to MCP server
    → Tools are discovered
    → Agent calls tools during the turn
    → Calls are logged in the thread timeline

Registration

MCP servers are registered at the tenant level, then assigned to agent templates or agents.

Typical config includes:

Field	Description
Name	Display name for the MCP server
URL	HTTP endpoint for the server
Transport	Streamable HTTP or SSE
Auth	None, tenant API key, or per-user OAuth

In the admin experience, this should appear under Connectors → MCP Tools.

Authentication

MCP tool connectors commonly use one of three patterns:

None

Useful for private-network services or systems with external network controls.

Tenant API key

A shared secret is stored securely and resolved at invocation time for any agent allowed to use that MCP server.

Per-user OAuth

Useful when downstream tools need user-scoped access. ThinkWork resolves the paired user’s token before opening the MCP connection.

Note

If a per-user OAuth token is missing, ThinkWork can skip that MCP server for the invocation while still allowing the rest of the agent run to proceed.

Runtime behavior

At runtime, ThinkWork discovers tools from the MCP server and adds them beside built-in tools and skill-provided tools. If an MCP server is unavailable, the agent can continue with its remaining toolset.

MCP usage is tracked in the thread timeline, including:

• tool name
• MCP server attribution
• input and output previews
• execution status

Supported use cases

Common MCP tool connector use cases include:

• CRM lookups and updates
• internal admin APIs
• task and ticket actions
• business-system queries
• search or retrieval over external systems

Related pages

• Connectors Overview
• Integrations