> This page is for Tools.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.moveworks.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.moveworks.com/_mcp/server.

# Tool Frameworks

A **tool framework** is a way of connecting the Moveworks AI Assistant to a business system or content source. Each framework is optimized for a different kind of connection: finding content, writing to an enterprise application, reading and making quick updates in a productivity app, or handing a task to another agent. Forms, a Service Management capability, is documented under [Service Management](/service-management/forms-assistant). The framework you choose for a use case affects how reliable, secure, and cost-effective the resulting agent is, so match it to the kind of work rather than to the system.

As a rule of thumb: use **Search** to find information, **MCP** to read and take reversible actions in productivity apps, **Plugins** for write actions that must respect business rules, and **Agent-to-Agent (A2A)** to hand off open-ended work to another agent.

#### [Search](/tool-studio/enterprise-search/search-connectors)

Use it for searching content and files, such as ServiceNow knowledge base articles and attachments.

#### [Plugins](/tool-studio/conversation-process)

Use it for reliable write actions, from straightforward requests to multi-step processes with real business logic.

#### [MCP Workspace](/tool-studio/mcp-workspace/overview)

Use it for live fetches against productivity apps and reversible operations, such as creating a Linear issue or fetching a Slack thread.

#### [Agent-to-Agent (A2A)](/tool-studio/a2a/connect-to-other-ai-agents)

Use it for open-ended, creative tasks that are best delegated to an agent with more latitude.

#### [Decision guide: Choosing the Right Tool Framework](/tool-studio/cookbooks/choosing-the-right-tool-framework)

The full comparison, with the trade-offs, limits, and examples behind each recommendation.

## Search

Enterprise Search hybridizes indexed and live search. Indexed search mirrors your content into a managed search index for fast, ranked retrieval across knowledge base articles, documents, and files; live search queries a connector's search API in real time for fast-moving content such as Slack threads. Moveworks maintains the [built-in connectors](/tool-studio/enterprise-search/search-connectors), which are configured through either Classic Ingestion or the Max Capacity journey depending on the content system. When no built-in connector covers a source, build one with [Content Gateway](/tool-studio/enterprise-search/content-gateway). For what Enterprise Search is and how it works, see the [Enterprise Search Overview](/ai-assistant/enterprise-search/overview).

Search administration lives here in Tools: [how content is ingested](/tool-studio/enterprise-search/content-ingestion-platform), [Classic Ingestion setup](/tool-studio/enterprise-search/classic-ingestion-setup), [Max Capacity setup](/tool-studio/enterprise-search/max-capacity-setup), [Profile Boosting](/tool-studio/enterprise-search/profile-boosting), [Vetted Content](/tool-studio/enterprise-search/vetted-content), [Permissions](/tool-studio/enterprise-search/permissions), and [Monitor Indexed Content](/tool-studio/enterprise-search/monitor-indexed-content).

## Plugins

Plugins deliver write actions against enterprise applications, primarily through their REST APIs. Automating a web app involves more than calling an API: forms expect fields in a specific order, business rules can live in the browser rather than the API, and dependent pick lists change which values are valid. Plugins handle this with [slot resolvers, policy validators, and the action orchestrator](/tool-studio/why-plugins), which enforce the order and rules a person would follow.

You design a plugin's behavior in its [Conversation Process](/tool-studio/conversation-process): [slots](/tool-studio/conversation-process/slots) collect inputs, [resolver strategies](/tool-studio/conversation-process/resolver-strategies) turn natural language into data, [activities](/tool-studio/conversation-process/activities) run actions and show content, and [control flow](/tool-studio/conversation-process/control-flow) branches on results. Build it by hand, or describe it to [ToolSmith](/tool-studio/toolsmith) and review the plan it proposes.

## MCP Workspace

The Model Context Protocol (MCP) connects the assistant to applications that already expose their tools through an MCP server. It fits read operations and reversible actions, where speed of connection matters more than heavy guardrails. [MCP Workspace](/tool-studio/mcp-workspace/overview) is where an admin connects and governs those servers: review sanctioned servers, see every tool a server exposes, and control which users can reach it. Start with [Connect and Manage Servers](/tool-studio/mcp-workspace/connect-and-manage-servers), validate with [Configure and Test MCP Workspace](/tool-studio/mcp-workspace/configure-and-test-mcp-workspace), inspect calls in [MCP tool call logs](/tool-studio/mcp-workspace/view-mcp-tool-call-logs), or [build a custom MCP server](/tool-studio/mcp-workspace/build-a-custom-mcp-server) for a private system.

MCP is less suited to high-stakes, multi-step write workflows that must follow a strict sequence, enforce business rules outside the API, or complete without partial failures. Use a plugin for those.

## Agent-to-Agent (A2A)

A2A connects Moveworks to other agents and existing AI investments. Agents that support A2A publish an Agent Card, and Moveworks calls them through the A2A protocol; applications that do not publish one are connected through plugins instead. A2A allows more variability than the other frameworks, so reserve it for cases where that flexibility is an advantage. [A2A Agents](/tool-studio/a2a) explains how a registered agent becomes a conversation the Agentic Orchestrator can hold, and [Connect an A2A Agent](/tool-studio/a2a/connect-an-a2a-agent) walks through registering one. On the Agentic Reasoning Engine, or for agents without an Agent Card, [A2A in the Agentic Reasoning Engine](/tool-studio/a2a/connect-to-other-ai-agents) walks through connecting an agent with a plugin.

> **Productivity Assistants**
>
> Productivity Assistants run on the Agentic Orchestrator, where Tool Studio plugins, MCP servers from MCP Workspace, A2A agents, and built-in tools all reach the model through one interface, scoped per assistant. [Tool Orchestration](/harness/agentic-orchestrator/tool-orchestration) explains how the Orchestrator presents each source and how confirmation and consent work.

## The Reverse Direction

The frameworks above describe how Moveworks connects out to your systems. The [Conversations API](/api-reference/conversations-api) is the reverse: it lets another agent or application call Moveworks as a tool, so your assistant's capabilities are available inside another AI framework.