Skip to main content

MCP File Download

Source example: agentflow/examples/github-mcp/mcp_file_download.py

What you will build​

A GitHub-MCP-backed agent that asks a remote MCP server to fetch a file from a repository, then uses the result inside the normal AgentFlow message loop.

Prerequisites​

  • Python 3.11 or later
  • 10xscale-agentflow installed
  • fastmcp installed
  • GITHUB_TOKEN
  • a model key such as GEMINI_API_KEY

Why this is a separate tutorial​

github-mcp.md focuses on repository metadata such as commits. This page focuses on file retrieval, which is a common pattern for:

  • reading README.md
  • fetching config files
  • summarizing source files

Tool flow​

Step 1 — Configure the remote GitHub MCP server​

The config is the same as the repository example:

config = {
    "mcpServers": {
        "github": {
            "url": "https://api.githubcopilot.com/mcp/",
            "headers": {"Authorization": f"Bearer {os.getenv('GITHUB_TOKEN')}"},
            "transport": "streamable-http",
        },
    }
}

Step 2 — Create the MCP-backed ToolNode and graph​

The graph structure is unchanged:

client_http = Client(config)
tool_node = ToolNode(tools=[], client=client_http)

main_agent = Agent(
    model="gemini-2.0-flash",
    provider="google",
    system_prompt=[...],
    tools=tool_node,
    trim_context=True,
)

The important difference is the prompt you send.

Step 3 — Ask for a repository file​

The example requests README.md:

inp = {
    "messages": [
        Message.text_message(
            "Get Readme.md file form the github repo "
            "'https://github.com/suchith83/portfolio' of the 'suchith83' username,."
        )
    ]
}
config = {"thread_id": "12345", "recursion_limit": 10}

res = app.invoke(inp, config=config)

Because the agent has access to GitHub MCP tools, it can choose the appropriate remote file-access tool and bring the result back into the message history.

File retrieval sequence​

Logging and debugging​

This example enables more logging than the other GitHub MCP page:

logging.basicConfig(level=logging.INFO)
logging.getLogger("agentflow").setLevel(logging.DEBUG)

That is helpful when debugging:

  • tool discovery
  • remote tool invocation
  • file retrieval failures

Verification​

Successful behavior should include:

  • a tool call in message history
  • a tool result related to repository file retrieval
  • a final assistant message that references README content

Common mistakes​

  • Missing or invalid GITHUB_TOKEN.
  • Assuming the file path exactly matches what the remote tool expects.
  • Forgetting that remote tools may return structured data rather than plain text.
  • Treating MCP file retrieval as a local filesystem operation.

Key concepts​

ConceptDetails
file-oriented MCP toolRemote tool that fetches repository file content
debug loggingHelps inspect remote tool behavior
same ReAct loopOnly the task and remote tool change, not the graph shape

What you learned​

  • How to use MCP for file retrieval workflows.
  • How to debug remote file-oriented tool calls.
  • How AgentFlow can summarize or transform file content after retrieval.

Next step​

→ Memory to add long-term user memory to a graph.