Resources

Resources is the working area for the reusable building blocks behind agents and conversations. If agents are the assistants, resources are the things those assistants rely on.

This page is the summary view. For a tour of one resource type — its screens, actions, and good habits — continue with the detailed guides:

You usually come here when you are preparing an agent, improving an existing setup, or checking how a workflow gets its data and actions.

The resource types

Knowledge bases store documents and indexed content that agents can search, so answers are grounded in trusted material. Content can be uploaded directly or kept in sync from external sources such as websites, Notion, cloud storage, and GitHub.
Prompts are reusable text instructions that keep wording consistent across chat requests, agent behavior, and recurring workflows.
Skills are reusable instruction sets — a packaged behavior or method — that can be attached to agents. A skill can be authored directly or fetched from a source URL.
MCP servers expose external tools to XTM One in a controlled way. Use them when agents need capabilities such as search, data lookups, or system actions provided through MCP.
Tools are the actions agents can use directly. In standard XTM One this is mainly built-in tools (image generation, code interpreter, web fetch, browser, and the platform tool); some deployments also allow custom tools.
Variables store reusable values that resolve at runtime in prompts, personas, skills, and tools, so key values are not repeated by hand. Values can be marked Secret.
Integrations connect XTM One to external services such as email, ticketing, messaging, and cloud document systems, giving agents live access to those systems.

Grid and list views

Every resource list page — and the Agents page — can be displayed two ways. The toggle sits at the right of the filter row, next to the item count.

Grid is the default: one card per item, optimized for browsing and recognizing things at a glance.
List shows the same items as table rows. It fits far more per screen and exposes extra attributes side by side, so you can compare a whole fleet at once — which model each agent runs, which tools are enabled, how many documents each knowledge base holds.

On the Agents page, favorites (the star on a card or row) sit above every other item in both views — sorting a column reorders the favorites among themselves rather than sending them back into the pile.

In list view, click a column header to sort by it; click again to reverse the order. Items with no value for that column (no model pinned, never fetched) always stay at the bottom, in both directions. Every action available on a card — opening the item, enabling or disabling it, starting a chat, copying a variable reference — is available on the matching row.

Your choice is remembered per page and across sessions, so a page you prefer as a table stays a table the next time you sign in.

Platform MCP servers

XTM One is not only an MCP client — it is also an MCP server for the connected Filigran platforms. As soon as an OpenCTI, OpenAEV, or OpenGRC instance is registered with XTM One, external MCP clients (Cursor, Claude Desktop, and others) can connect to /mcp/opencti, /mcp/openaev, and /mcp/opengrc with a personal API key and work with threat intelligence, exposure validation, and governance-risk-compliance content directly.

See Platform MCP servers for endpoints, authentication, the tool surface, and client configuration.

Platform tools MCP server

XTM One also exposes itself over MCP, on /mcp/xtm: agent traces, agents, assignments, runs, knowledge bases, prompts, skills, custom tools, integrations, and variables — scoped to what the calling user can already see and manage. It is how you read an agent's traces from Claude Desktop or Cursor and fix its configuration without leaving the client. Off by default; an administrator enables it in Settings > MCP Endpoint.

See Platform tools MCP server.

How the resource types work together

In a typical workflow:

an integration connects to the outside system
a knowledge base provides trusted context
a prompt or skill shapes the behavior
a tool performs an action
a variable avoids repeating key values

This is why Resources matters so much. It is the place where assistants become usable and consistent.

Next step

Continue with the detailed guide of the resource type you need, starting with Integrations.