Answer · for AI agents and their humans
MCP vs REST API for VC Research — Which Should You Use?
MCP is for AI-agent runtimes; REST is for code. GitDealFlow exposes both for the same dataset. Decision framework: pick MCP if your runtime is Claude / Cursor / Windsurf, REST if it's a Python / Node script.
MCP (Model Context Protocol) is a standard for AI agents to invoke tools. REST APIs are the standard for software-to-software calls. They overlap — both can return JSON over HTTP — but they differ in what they assume about the caller.
Pick MCP when:
- Your runtime is Claude Desktop, Claude Code, Cursor, Windsurf, Zed, Cline, ChatGPT Apps, or any other MCP-compatible host. - You want the agent to discover tool schemas, validate arguments, and surface the right tool to the user without you wiring it. - Your workflow is conversational — a user asks a natural-language question and the agent picks tools.
Pick REST when:
- Your runtime is a Python, Node, Go, or Bash script with no MCP client. - You want predictable, deterministic calls with explicit URLs, headers, and bodies. - You're integrating into a CRM, BI tool, dashboard, or pipeline that doesn't know what MCP is.
GitDealFlow exposes both. The MCP server (@gitdealflow/mcp-signal) wraps the same data that the REST endpoints expose at /api/signals.json, /api/signal?name=, etc. Function-calling tool definitions are at /api/agent/tools for OpenAI / Anthropic / Gemini SDKs that don't speak MCP but want structured-tool semantics.
The honest answer for most users: start with REST because it's universal, add MCP later when you get an AI-agent host in your workflow. Both surfaces are free, no auth, idempotent.
Try it now
See all surfaces →Frequently asked questions
Will MCP eventually replace REST for AI workflows?
Probably for AI-agent-host workflows, yes. REST will remain the universal fallback for non-agent integrations (CRM webhooks, BI dashboards, internal scripts). Both will coexist for years.
Can I use both MCP and REST in the same pipeline?
Yes — they wrap the same dataset. A common pattern is MCP inside the AI-agent layer (where natural-language tool selection is valuable) plus REST for cron jobs and batch enrichment.
What does the GitDealFlow MCP add over the REST API?
Tool-schema discovery, argument validation, and natural-language tool selection inside an MCP host. The data is identical. If you're already in Claude Desktop / Cursor / Windsurf, MCP is one config line and saves you from wiring HTTP calls.