docs(TSP-1166): document MCP read_resource capability

[claude]

Summary

• Updated integrations/mcp/mcp-client.mdx to document the new read_resource capability introduced in PR #13574
• Added a new Understanding MCP capabilities section explaining the distinction between tools (model-controlled, perform actions) and resources (application-controlled, read-only context)
• Updated setup section to mention agents get both tools and resources on connection
• Added a new Using MCP resources section with concrete examples (documentation files, skill guides, schemas, API references)
• Added a resource-specific troubleshooting accordion
• Added two new FAQs: tools vs resources difference, and whether agents auto-read resources

Related

Linear: https://linear.app/relevance/issue/TSP-1166/

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
relevanceai	🟢 Ready	View Preview	Apr 17, 2026, 5:41 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[linear]

TSP-1166 [Frontend] PR #13574: Agents can now read MCP resources (not just tools)

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🟢	Consistency	9/10	One stale internal link using a redirect path instead of the canonical destination. No banned words, no British spellings, capitalization is solid throughout.
🟡	Technical clarity	8/10	The internal link on line 10 points to a redirect path rather than the canonical URL — fine for now but will silently break if the redirect is ever cleaned up.
🟡	Non-technical clarity	7/10	MCP is used throughout without ever being expanded or defined. A reader new to the concept has no anchor.
🟢	Structure	9/10	Numbered procedure on lines 21–26 would read better as a <Steps> component. Everything else flows well.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: Solid, well-scoped page — clear intro, good gotchas (transport requirement, local server limitation, (MCP: connection-name) label behavior), and the troubleshooting/FAQ structure is appropriate. Two small fixes and one component swap would get this to 10 across the board.

🔧 Issues (2)

• integrations/mcp/mcp-client.mdx:10 — Link target /integrations/mcp/programmatic-gtm/introduction is a redirect (docs.json line 818–819 maps it to /get-started/core-concepts/programmatic-gtm). Use the canonical destination directly so the link survives any future redirect cleanup: /get-started/core-concepts/programmatic-gtm or /integrations/mcp/programmatic-gtm/mcp-server (the latter is probably what a user following the "use Relevance AI from Claude or Cursor" path actually needs).

• integrations/mcp/mcp-client.mdx (whole file) — "MCP" is used from the first sentence without expansion or definition. The sibling page programmatic-gtm/mcp-server.mdx defines it in an FAQ accordion: "The Model Context Protocol (MCP) is an open standard…". Expand the acronym at first use here too — either inline ("MCP (Model Context Protocol) server") or via a brief parenthetical in the intro paragraph.

🧩 Component suggestions (1)

• integrations/mcp/mcp-client.mdx:21–26 — The four-step numbered list under "Setting up an MCP connection for Agents" is a sequential procedure. CLAUDE.md recommends <Steps> for these since it adds visual progress indicators. Wrap as:

  <Steps>
    <Step title="Open your Agent">
      Open the Agent you want to connect.
    </Step>
    <Step title="Navigate to Tools">
      Go to the **Tools** tab in the Agent builder.
    </Step>
    <Step title="Add MCP">
      Click **Add MCP**.
    </Step>
    <Step title="Choose your connection type">
      Select a preset (Notion, Canva, or Atlassian) and follow the OAuth flow, or click **Connect my own** to add a custom server URL, label, and credentials.
    </Step>
  </Steps>

⚠️ Contradictions (0)

No contradictions found. The mutual cross-links between mcp-client.mdx and programmatic-gtm/mcp-server.mdx are consistent — each correctly describes the other's scope and links appropriately.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	2
Total lines processed	~428

Files read: integrations/mcp/mcp-client.mdx (94 lines), integrations/mcp/programmatic-gtm/mcp-server.mdx (334 lines)