A2A INTEGRATION · CUSTOM TOOL

Use GitDealFlow A2A with OpenAI Agents SDK

openai-agents Python and TypeScript packages. Builders shipping agent workflows on OpenAI models with first-class tool use.

Crunchbase API: $20K/yr. GitDealFlow A2A: free, no signup.

The OpenAI Agents SDK gives you a clean primitive for tool use. A2A is not natively supported by the SDK as of April 2026, so the integration is a custom function tool that POSTs JSON-RPC to our endpoint. The whole tool fits in 30 lines and exposes all five skills behind a single dispatch function.

Install

pip install openai-agents
# or
npm install @openai/agents

Custom function tool (Python)

# pip install openai-agents requests
from openai_agents import Agent, function_tool
import requests

A2A_URL = "https://signals.gitdealflow.com/api/a2a"

@function_tool
def gitdealflow_query(skill: str, args: dict | None = None) -> dict:
    """
    Call the GitDealFlow A2A agent.
    skill: one of get_trending_startups, search_startups_by_sector,
           get_startup_signal, get_signals_summary, get_methodology
    args:  skill-specific arguments (e.g. {"sector": "fintech"})
    """
    body = {
        "jsonrpc": "2.0", "id": 1,
        "method": "message/send",
        "params": {
            "message": {
                "role": "user",
                "parts": [{"kind": "data", "data": {"skill": skill, "args": args or {}}}],
            }
        },
    }
    r = requests.post(A2A_URL, json=body, timeout=15)
    r.raise_for_status()
    return r.json()["result"]["artifacts"][0]["parts"][0]["data"]

agent = Agent(
    name="VC scout",
    tools=[gitdealflow_query],
    instructions="Use gitdealflow_query to fetch live startup engineering signals.",
)

TypeScript custom tool

// npm install @openai/agents
import { Agent, tool } from "@openai/agents";
import { z } from "zod";

const A2A_URL = "https://signals.gitdealflow.com/api/a2a";

const gitdealflow = tool({
  name: "gitdealflow_query",
  description: "Call the GitDealFlow A2A agent for startup engineering signals.",
  parameters: z.object({
    skill: z.enum([
      "get_trending_startups",
      "search_startups_by_sector",
      "get_startup_signal",
      "get_signals_summary",
      "get_methodology",
    ]),
    args: z.record(z.string(), z.any()).optional(),
  }),
  execute: async ({ skill, args }) => {
    const r = await fetch(A2A_URL, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({
        jsonrpc: "2.0",
        id: 1,
        method: "message/send",
        params: {
          message: {
            role: "user",
            parts: [{ kind: "data", data: { skill, args: args ?? {} } }],
          },
        },
      }),
    });
    const json = await r.json();
    return json.result.artifacts[0].parts[0].data;
  },
});

What you can ask

Pass {skill: 'get_trending_startups'} for weekly top 20.
Pass {skill: 'search_startups_by_sector', args: {sector: 'ai-ml'}}.
Pass {skill: 'get_startup_signal', args: {name: 'Roboflow'}}.
Pass {skill: 'get_methodology'} when an analyst questions the ranking.

Gotchas

The endpoint returns JSON-RPC envelopes. Always reach into result.artifacts[0].parts[0].data for the structured payload, or .parts[1].text for the plaintext fallback.
Sector slugs are normalized aliases (crypto → web3, AI → ai-ml). If you accept user free-text, route it through search_startups_by_sector and let the endpoint resolve.

References

Try it without setup

The interactive playground lets you send live JSON-RPC requests against the A2A endpoint with no install, no auth. Pick a skill, hit send, see the response.

Full launch story: I made my VC deal flow callable by Claude.

Other framework integrations

Custom function tool (Python)

# pip install openai-agents requests from openai_agents import Agent, function_tool import requests A2A_URL = "https://signals.gitdealflow.com/api/a2a" @function_tool def gitdealflow_query(skill: str, args: dict | None = None) -> dict: """ Call the GitDealFlow A2A agent. skill: one of get_trending_startups, search_startups_by_sector, get_startup_signal, get_signals_summary, get_methodology args: skill-specific arguments (e.g. {"sector": "fintech"}) """ body = { "jsonrpc": "2.0", "id": 1, "method": "message/send", "params": { "message": { "role": "user", "parts": [{"kind": "data", "data": {"skill": skill, "args": args or {}}}], } }, } r = requests.post(A2A_URL, json=body, timeout=15) r.raise_for_status() return r.json()["result"]["artifacts"][0]["parts"][0]["data"] agent = Agent( name="VC scout", tools=[gitdealflow_query], instructions="Use gitdealflow_query to fetch live startup engineering signals.", )

TypeScript custom tool

// npm install @openai/agents import { Agent, tool } from "@openai/agents"; import { z } from "zod"; const A2A_URL = "https://signals.gitdealflow.com/api/a2a"; const gitdealflow = tool({ name: "gitdealflow_query", description: "Call the GitDealFlow A2A agent for startup engineering signals.", parameters: z.object({ skill: z.enum([ "get_trending_startups", "search_startups_by_sector", "get_startup_signal", "get_signals_summary", "get_methodology", ]), args: z.record(z.string(), z.any()).optional(), }), execute: async ({ skill, args }) => { const r = await fetch(A2A_URL, { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ jsonrpc: "2.0", id: 1, method: "message/send", params: { message: { role: "user", parts: [{ kind: "data", data: { skill, args: args ?? {} } }], }, }, }), }); const json = await r.json(); return json.result.artifacts[0].parts[0].data; }, });

Gotchas

The endpoint returns JSON-RPC envelopes. Always reach into result.artifacts[0].parts[0].data for the structured payload, or .parts[1].text for the plaintext fallback.

Sector slugs are normalized aliases (crypto → web3, AI → ai-ml). If you accept user free-text, route it through search_startups_by_sector and let the endpoint resolve.