BrainBank
AI Classroom/MCPClaude Code Deep Dive

ListMcpResourcesTool: List MCP resources

8/2/2026, 4:54:01 PM · Source

AI-translated on 8/2/2026, 4:54:56 PM · by Qwen3.6 35B (fast, default)

#mcp#resource-discovery#listmcpresources#python-sdk#model-context-protocol

ListMcpResourcesTool is a resource discovery tool in the MCP ecosystem that enables models to dynamically list all available resource URIs after connecting to an MCP Server, avoiding blind guessing.

It Grants the Model "Resource Discovery Capability" First

ListMcpResourcesTool's role is much like directory browsing in a remote environment.
If the model has already connected to an MCP server but doesn't know exactly what resources are exposed there, the first step should be to use it to list them out.

Key Source Code

tools/ListMcpResourcesTool/ListMcpResourcesTool.ts:

const inputSchema = z.object({
  server: z.string().optional().describe('Optional server name to filter resources by'),
})

The core logic is:

const results = await Promise.all(
  clientsToProcess.map(async client => {
    const fresh = await ensureConnectedClient(client)
    return await fetchResourcesForClient(fresh)
  }),
)

This shows that it doesn't rely on reading a static cache, but instead operates around the MCP client connection state.

Call Chain

Model needs to read MCP resource > ListMcpResourcesTool > Verify connected client > fetchResourcesForClient > Return resource list > Pass to ReadMcpResourceTool

Summary

ListMcpResourcesTool addresses the resource discovery problem; without it, the model would have to blindly guess URIs.

Learning map

🗺️ Learning Roadmap: Mastering MCP Resource Discovery and Reading

Phase 1: Foundational Knowledge

  • Understand MCP (Model Context Protocol) core concepts — Resources, Tools, Prompts
  • Understand Resource URI formats and classifications (file://, database://, custom protocols, etc.)
  • Clarify the position of resource discovery within the Agent workflow

Phase 2: Environment Setup

  • Install the MCP Python/TypeScript SDK
  • Configure and start an MCP Server (example: filesystem server)
  • Verify bidirectional Client ↔ Server connection status

Phase 3: Using ListMcpResourcesTool

  • Call list_resources() / resources/list method to retrieve the resource list
  • Use optional parameters to filter resources by server name
  • Understand the return result structure (URI, MIME type, description, and other metadata)

Phase 4: In-Depth Practice

  • Combine ReadMcpResourceTool to complete the full "discover → read" chain
  • Handle pagination with cursors (cursor) to retrieve all resources
  • Manage client connection pools in multi-Server scenarios

Phase 5: Advanced Exploration

  • Write a custom MCP Server and expose ResourceTemplate
  • Dynamically register mutable virtual resources (non-static files)
  • Integrate resource browsing UI into Agent IDEs such as Claude Code / Cursor

Get hands-on — step by step

  1. Install the MCP SDK: pip install mcp or npm install @modelcontextprotocol/sdk
  2. Initialize an MCP Client and connect to the server: 
    from mcp import ClientSession
session = await ClientSession()
await session.connect(
    subprocess.Popen(["uvicorn", "my_server:app"], stdin=subprocess.PIPE, stdout=subprocess.PIPE)
)
  3. Call the resource list endpoint; omit parameters to fetch all: 
    resources = await session.list_resources()
for r in resources:
    print(f"{r.uri} — {r.mimeType or 'text'}")
  4. Filter by server name (if needed): Filter the resources result beforehand, or implement via a server filter on the Server side.
  5. (Optional) Handle pagination: If the response includes nextCursor, pass that cursor to call list_resources(cursor=...) again until there are no more pages.
  6. Combine with a read operation: Select a URI from the list and pass it to ReadMcpResourceTool: 
    content, mime_type = await session.read_resource("file:///path/to/file.txt")
print(content)
  7. Verify the end-to-end flow: Configure the MCP Server in Claude Code / Cursor and observe the complete call chain as the Agent automatically invokes resource discovery and reading.
  8. Explore custom servers: Create a new server that exposes a ResourceTemplate (dynamic URI prefix), restart the client, and repeat steps 3-6.

Top 3 sources

  1. 1
    MCP 官方规范 - Resources

    MCP 协议 specification，详细描述 resources/list、资源 URI 格式及分页机制。

    https://modelcontextprotocol.io/specification/2025-06-18/server/resources

  2. 2
    Anthropic MCP Python SDK 文档

    MCP 官方 Python SDK，包含 list_resources()、read_resource() 等核心 API 的用法与示例。

    https://github.com/modelcontextprotocol/python-sdk

  3. 3
    Phil Schmidt：MCP 入门指南

    一篇清晰的 MCP 概览文章，解释了资源发现（Discovery）、上下文提供和 Agent 工作流中 MCP 的位置。

    https://www.philschmid.de/mcp-introduction

Links are AI-suggested — worth a quick sanity check before diving in.