For AI Agents

All programmatic surfaces for autonomous agents, LLM tool-calling, and crawler discovery

Last updated: July 30, 2026

VC Deal Flow Signal exposes 8 programmatic surfaces designed for AI agents, LLM tool-calling, and autonomous crawlers. Every surface is free, public, and requires no authentication. Use these endpoints to integrate startup engineering velocity data into your agent, copilot, or crawler pipeline.

Programmatic Surfaces

1️⃣ MCP Server — Stdio Transport
6 tools for agentic access: get trending startups, search by sector, get startup signal, scout score, methodology, and summary. Works with Claude Desktop, Claude Code, Cursor, and any MCP-compatible host.
npx -y @gitdealflow/mcp-signal
Claude Desktop / Claude Code config:
{
  "mcpServers": {
    "vc-deal-flow-signal": {
      "command": "npx",
      "args": ["-y", "@gitdealflow/mcp-signal"]
    }
  }
}
2️⃣ MCP Server — Streamable HTTP Transport
Same 6 tools via HTTP POST for ChatGPT Apps, hosted MCP clients, and any HTTP MCP runtime.
POST https://signals.gitdealflow.com/api/mcp/rpc
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "get_trending_startups",
    "arguments": {}
  },
  "id": 1
}
🔗 JSON-RPC methods: initialize, tools/list, tools/call, resources/list, resources/read, prompts/list, prompts/get
3️⃣ A2A Endpoint (Google Agent-to-Agent)
Google A2A-compatible JSON-RPC 2.0 endpoint (protocolVersion 0.3.0). Five skills map to the MCP tools. Designed for Google A2A agents and orchestrators.
POST https://signals.gitdealflow.com/api/a2a
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "method": "tasks/send",
  "params": {
    "id": "trending-1",
    "skillId": "get_trending_startups",
    "input": {}
  }
}
4️⃣ NLWeb Endpoint (Microsoft Bing Copilot)
Natural-language query interface. Accepts conversational queries, returns schema.org-typed JSON-LD answers (ItemList, Organization, Article, Dataset, WebPage). Compatible with Bing Copilot and Microsoft NLWeb-aware crawlers.
curl -X POST https://signals.gitdealflow.com/api/nlweb \
  -H "Content-Type: application/json" \
  -d '{"query": "trending fintech startups this week"}'
Supported intents: trending, <sector> startups, tell me about <startup>, methodology, summary.
🔗 GET /api/nlweb returns the descriptor with example queries and schema.
5️⃣ Function-Calling API (OpenAI / Anthropic / Gemini)
Provider-native tool definitions. Drop directly into OpenAI, Anthropic Claude, or Google Gemini SDKs without needing an MCP client.
# Get tool definitions
curl https://signals.gitdealflow.com/api/agent/tools?format=openai
curl https://signals.gitdealflow.com/api/agent/tools?format=anthropic
curl https://signals.gitdealflow.com/api/agent/tools?format=gemini

# Invoke a tool
curl -X POST https://signals.gitdealflow.com/api/agent/call \
  -H "Content-Type: application/json" \
  -d '{"name": "get_trending_startups", "arguments": {}}'
🔗 GET /api/agent/tools (no query) returns all formats keyed by provider plus the invoke URL.
6️⃣ JSON API
Full dataset as JSON. 313 startups, 15 sectors, commit velocity, contributor growth, signal types, and metadata. Ideal for direct HTTP consumption or AI SDK tool calls.
curl https://signals.gitdealflow.com/api/signals.json
🔗 signals.json (142KB, updated weekly)
7️⃣ CSV Export
Same data as JSON in CSV format. Perfect for spreadsheets, dataframes, BI tools, and batch processing.
curl https://signals.gitdealflow.com/api/signals.csv
8️⃣ OpenAPI 3.1 Specification
Complete OpenAPI 3.1.0 specification for all API endpoints. Use for code generation, tool registries, and API documentation.
curl https://signals.gitdealflow.com/api/openapi.json

MCP Tools Reference

All six MCP tools are read-only and idempotent. No parameter validation needed beyond what's in the input schema.
ToolDescription
get_trending_startupsTop 20 startups across all sectors
search_startups_by_sector(sector)Filter by sector slug (15 enumerated values)
get_startup_signal(name)Case-insensitive startup profile lookup
get_signals_summaryPeriod, freshness, format URLs
get_scout_receipts(github_username)Scout Score (0-100) from starring history vs. 75+ unicorns
get_methodologyFull methodology + canonical URL

Agent Discovery Files

The following machine-readable discovery files are available for autonomous agent crawling and LLM registration:
FilePurpose
mcp.jsonMCP server discovery manifest
agent-card.jsonA2A AgentCard (protocolVersion 0.3.0)
ai-plugin.jsonChatGPT plugin manifest
agents.jsonAgents.json discovery standard
openai-search.jsonOpenAI search crawler manifest
AGENTS.mdAgent reference (markdown)
llms.txtLLM page index (gitdealflow.com)
llms-full.txtFull agent reference