Make HTTP Requests to External APIs

RequestsToolSpec equips LlamaIndex agents with the ability to make HTTP requests (GET, POST, PATCH) to external APIs. It provides a security-first approach by requiring explicit hostname-to-header mappings, ensuring credentials are only sent to authorized domains. The tool can be combined with OpenAPIToolSpec to interface with OpenAPI servers, enabling agents to interact with RESTful services.

Use RequestsToolSpec when your agent needs to fetch data from REST APIs, submit form data, or update resources via HTTP methods. Combine it with OpenAPIToolSpec when working with OpenAPI-documented endpoints.

Avoid it when your security model cannot accommodate pre-configured domain headers, as the tool requires upfront hostname-to-credential mapping rather than dynamic authentication.

You provide a domain_headers dictionary mapping hostnames to their required HTTP headers (authorization tokens, content types, custom headers). The tool exposes three methods to your agent: get_request for retrieving data, post_request for creating resources, and patch_request for updating resources. Each method accepts a URL and returns the HTTP response, which the agent can parse and act upon.

RequestsToolSpec integrates with OpenAPIToolSpec to interface with OpenAPI servers, enabling schema-driven API interactions. It works within the LlamaIndex agent framework, specifically with FunctionAgent and compatible LLM providers like OpenAI models. The tool is designed as a ToolSpec that converts to a tool list consumable by LlamaIndex agents.

This tool serves developers building LlamaIndex agents that need to interact with external HTTP APIs.

from llama_index.tools.requests import RequestsToolSpec
from llama_index.core.agent.workflow import FunctionAgent
from llama_index.llms.openai import OpenAI

domain_headers = {
    "api.openai.com": {
        "Authorization": "Bearer sk-your-key",
        "Content-Type": "application/json",
    }
}

tool_spec = RequestsToolSpec(domain_headers=domain_headers)

agent = FunctionAgent(
    tools=tool_spec.to_tool_list(),
    llm=OpenAI(model="gpt-4.1"),
)

print(await agent.run("<query>"))