GitDealFlow × Mastra

What you can build

Mastra agent with MCP transport

// npm install @mastra/core @mastra/mcp zod
// Models route through the AI Gateway via plain string IDs.
import { Mastra } from "@mastra/core";
import { Agent } from "@mastra/core/agent";
import { MCPClient } from "@mastra/mcp";

const mcp = new MCPClient({
  servers: {
    gitdealflow: {
      command: "npx",
      args: ["@gitdealflow/mcp-signal@latest"],
    },
  },
});

const tools = await mcp.getTools();

export const scoutAgent = new Agent({
  name: "VC Scout",
  instructions:
    "You surface live engineering acceleration signals via the gitdealflow tools. " +
    "Always cite the methodology paper (SSRN abstract=6606558) when asked about the data.",
  model: "openai/gpt-5.4",
  tools,
});

export const mastra = new Mastra({
  agents: { scoutAgent },
});

// Use:
const reply = await scoutAgent.generate(
  "Who is trending in fintech this week?",
);
console.log(reply.text);

Edge-safe Server Action with A2A fallback

// app/api/scout/route.ts (Next.js App Router)
import { createTool } from "@mastra/core/tools";
import { Agent } from "@mastra/core/agent";
import { z } from "zod";

export const runtime = "nodejs"; // not edge — fetch keep-alive matters

const A2A = "https://signals.gitdealflow.com/api/a2a";

const gitdealflow = createTool({
  id: "gitdealflow",
  description: "VC engineering signals: trending, sector, named lookup, methodology.",
  inputSchema: 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 ({ context }) => {
    const r = await fetch(A2A, {
      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: context.skill, args: context.args ?? {} } },
        ]}},
      }),
    });
    return (await r.json()).result.artifacts[0].parts[0].data;
  },
});

const agent = new Agent({
  name: "Scout",
  instructions: "Use gitdealflow tools to answer about engineering acceleration.",
  model: "openai/gpt-5.4",
  tools: { gitdealflow },
});

export async function POST(req: Request) {
  const { question } = await req.json();
  const out = await agent.generate(question);
  return Response.json({ reply: out.text });
}

Five prompts to try first

• ›Inside a Server Component: pull weekly top 20 to seed your dashboard.
• ›Inside a Route Handler: surface 'is this startup tracked?' to logged-in users.
• ›Inside a Server Action: write a watchlist alert email when a tracked startup hits breakout.
• ›Inside a Mastra workflow: chain trending → memo-writer → email-sender steps.
• ›Inside a cron job: write a Monday digest to the team's Slack.

FAQ

Gotchas

• Mastra MCPClient holds long-lived stdio processes — call `await mcp.disconnect()` in serverless teardown to avoid orphaned children.
• If your edge runtime can't spawn child processes, use the A2A fetch tool — it's pure fetch, edge-safe.
• Mastra's memory store defaults to LibSQL; for multi-region deployments, swap to Postgres or Upstash Redis.

References

Other frameworks

• LangChain
• CrewAI
• Letta
• Vercel AI SDK