What's the difference between wmcp.sh and Pipedream?

Pipedream is a workflow automation platform — you build multi-step flows triggered by events (webhooks, schedules, queues), each step calls one of 2000+ pre-built integrations. wmcp.sh exposes any URL as MCP tools your AI agent can call directly. Pipedream is about chaining steps; wmcp.sh is about exposing surface for agents to drive. They serve different jobs.

When should I pick Pipedream?

When you're building backend workflows: 'when X happens, do Y then Z'. Their 2000+ integrations cover almost every SaaS. Their visual builder + free tier are excellent for non-agent flows. Pick Pipedream if you're triggering automations from events, not letting an agent reason about which tool to call next.

When should I pick wmcp.sh?

When you're building an AI agent that needs to call tools dynamically — Claude / Cursor / Codex deciding which MCP tool to invoke. wmcp.sh exposes ANY URL as MCP tools an agent can choose from. Pipedream's connectors are designed for human-defined workflows; they don't translate cleanly to agent tool_use.

Yes. Use Pipedream for backend event-driven workflows. Use wmcp.sh for the agent-facing tool layer. Many teams use Pipedream as a webhook receiver that feeds wmcp.sh-exposed APIs an agent then queries.

Pipedream: free 10k invocations/mo, then $19/mo for 1M invocations. wmcp.sh: free 100 reads/day anonymous, $29/mo Pro for 10k/day reads+executes. Different metering — Pipedream charges per workflow execution, wmcp.sh charges per tool call.

head-to-head · honest framing

wmcp.sh vs Pipedream.

Q: Does Pipedream do MCP?

Not as a primary product. They have some agent-related features but their core surface is workflow steps, not MCP tool_use shapes. wmcp.sh is MCP-native: every URL output is shaped for MCP clients (Claude.ai, Cursor, Codex, OpenCode, mcp-remote) directly.

Pipedream is a serverless workflow + integration platform with 2000+ connectors — built for backend event-driven automations. wmcp.sh is a different shape: URL → MCP tools for AI agents to call. Both legit, different jobs.

The shape difference

One sentence each.

Pipedream

A workflow automation platform — visual + code-first multi-step flows triggered by events (webhooks, schedules, queues). 2000+ pre-built integrations, generous free tier, mature platform.

wmcp.sh

A worker that turns any URL into MCP tools an AI agent can call dynamically. Agent-first surface — Claude / Cursor / Codex decide which tool to invoke at runtime.

Side-by-side

The capability matrix.

Capability	Pipedream	wmcp.sh
Trigger-based workflows "when X event happens, do Y"	✅ Core feature, mature	❌ Not the goal — wmcp.sh is request/response, not event-driven
Pre-built SaaS connectors	✅ 2000+ curated	⚠️ Covers via OpenAPI ingest, but no per-provider hand-tuning
Agent tool_use shape Claude / OpenAI function_call / MCP	⚠️ Some agent features; not MCP-native	✅ MCP / tool_use / function_call shapes, native
Shopper-side commerce	❌ Not in scope	✅ Shopify storefronts (4M+ stores) out of the box
Oracle / price-data CoinGecko, Pyth, Chainlink, DefiLlama	❌ Not in scope	✅ 5 adapters at /price-data
OAuth-proxy for MCP servers	❌ Not in scope	✅ RFC 7591 DCR + PKCE proxy at /mcp/<provider>
Free tier	10k invocations/mo	100 reads/day anonymous, no signup
Open source	Components open, platform closed	Worker + adapters MIT
Run on the edge	Centralized	Cloudflare Workers, 300+ POPs

Where each wins

The honest version.

Pipedream wins when:

You're building event-driven backend automations (webhook → process → forward)
You need 2000+ SaaS connectors hand-curated for stability
You want a generous free tier for non-agent workflow execution
You need a visual builder for non-technical operators

wmcp.sh wins when:

You're building an AI agent that decides which tool to call (vs. fixed workflow steps)
You need MCP-native — Claude.ai / Cursor / Codex / OpenCode plug into wmcp.sh directly
You need shopper-side, oracle, or DeFi adapters (Pipedream doesn't ship these)
You want OAuth-proxy so OAuth-gated MCP servers (DefiLlama, etc.) work in agents that can't drive OAuth themselves

Try wmcp.sh

One curl.

curl 'https://wmcp.sh/api/v1/tools?url=https://api.llama.fi/protocols'
# 6 DeFi MCP tools, agent-ready. No signup, no workflow builder.

FAQ

Common questions.

What's the difference?

Pipedream chains pre-defined workflow steps triggered by events. wmcp.sh exposes URLs as MCP tools an agent calls dynamically. Different jobs.

When pick Pipedream?

Event-driven backend flows. 2000+ connectors. Generous free tier.

When pick wmcp.sh?

Agent-facing tool layer. MCP-native. Shopper-side + OpenAPI ingest + oracle/DeFi.

Can I use both?

Yes. Pipedream for backend automations + wmcp.sh as the agent-facing tool layer.

Pricing?

Pipedream: free 10k invocations/mo, $19/mo for 1M. wmcp.sh: free 100/day anonymous, $29/mo Pro for 10k/day.

Does Pipedream do MCP?

Not as a primary product. wmcp.sh is MCP-native: every URL output is shaped for MCP clients directly.