Plug Dynamic Feed into any agent.

MCP clients, one config

Claude Desktop, Cursor, Cline, Windsurf and any MCP client. Add this and restart:

{
  "mcpServers": {
    "dynamic-feed": { "url": "https://dynamicfeed.ai/mcp" }
  }
}

Prefer a local command? (firewalls/VPNs that block remote streams)

npx -y dynamicfeed-mcp        # Node
uvx dynamicfeed-mcp           # Python

Use a focused MCP collection

Use /mcp when an agent needs the full public catalogue. For a narrower job, point the same client configuration at one centrally defined collection:

https://dynamicfeed.ai/mcp/security
https://dynamicfeed.ai/mcp/weather
https://dynamicfeed.ai/mcp/space
https://dynamicfeed.ai/mcp/infrastructure
https://dynamicfeed.ai/mcp/ai-models
https://dynamicfeed.ai/mcp/evidence

A smaller catalogue sends fewer tool descriptions into the model context. That can reduce prompt-token use and tool-selection ambiguity. Collections are filtered views of the same canonical registry, not separate copied servers; /mcp continues to expose the full public catalogue.

Frameworks

LangChain, langchain-mcp-adapters

from langchain_mcp_adapters.client import MultiServerMCPClient
client = MultiServerMCPClient({"dynamic-feed": {
    "url": "https://dynamicfeed.ai/mcp", "transport": "streamable_http"}})
tools = await client.get_tools()

LlamaIndex, llama-index-tools-mcp

from llama_index.tools.mcp import BasicMCPClient, McpToolSpec
spec = McpToolSpec(client=BasicMCPClient("https://dynamicfeed.ai/mcp"))
tools = await spec.to_tool_list_async()

Vercel AI SDK (v6), @ai-sdk/mcp

import { createMCPClient } from '@ai-sdk/mcp';
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js';
const transport = new StreamableHTTPClientTransport(new URL('https://dynamicfeed.ai/mcp'));
const mcp = await createMCPClient({ transport });
const tools = await mcp.tools();

OpenAI Agents SDK

from agents.mcp import MCPServerStreamableHttp
server = MCPServerStreamableHttp(params={"url": "https://dynamicfeed.ai/mcp"})

Pydantic AI

from pydantic_ai.mcp import MCPServerStreamableHTTP
server = MCPServerStreamableHTTP("https://dynamicfeed.ai/mcp")

ChatGPT / Custom GPT, Actions

In the GPT builder → Create new action → Import from URL, paste the agent-ready OpenAPI. Every operation is a GET flagged non-consequential (no confirmation prompts).

https://dynamicfeed.ai/openapi-agent.json

REST API

Prefer plain HTTP? The keyless batch sandbox accepts small, read-only test batches at /v1/batch. Documented direct REST routes use X-API-Key for keyed access.

curl -s https://dynamicfeed.ai/v1/batch \
  -H 'Content-Type: application/json' \
  -d '{"calls":[{"tool":"current_weather","args":{"city":"Tokyo"}}]}'