Home / Compare / Helve vs run.pay

Helve vs run.pay — Pay-Per-Call API & MCP Gateway for AI Agents

Both Helve and run.pay (getrunpay.com) give AI agents pay-per-call access to third-party APIs without juggling vendor accounts. Helve goes further with 313+ live tools across 255+ providers, 291+ free public APIs at $0.00/call, native Streamable HTTP MCP at https://helve.dev/mcp, and multi-provider fusion search.

CapabilityHelve (helve.dev)run.pay (getrunpay.com)
Catalog Breadth313+ live tools across 255+ providers205 services
Pricing & Free APIs291+ free public APIs at $0.00/call; metered at upstream + 15% ($0.50 free credit)$0.005+/call + 2% platform fee
MCP IntegrationNative Streamable HTTP MCP at https://helve.dev/mcpHosted via Smithery
Missing-Tool OnboardingBuilt-in request_tool & POST /v1/tools/request auto-onboardingStatic catalog
Failed Call Billing$0.00 debited when upstream provider failsStandard per-call fee model
Per-Key Spend GuardrailsDeterministic spend_limit_usd per sk_live_ agent keyAccount-level balance
Multi-Provider SearchBuilt-in provider: "fusion" consensus search across 5 enginesSingle-provider pass-through

Get API Key ($0.50 Free) Browse 313 Tools

FAQ

How does Helve compare to run.pay (getrunpay.com) for AI agents?

Helve serves 313+ live tools across 255+ providers (including 291+ free public APIs at $0.00/call), whereas run.pay lists 205 services starting at $0.005+/call plus a 2% fee. Helve also provides a native Streamable HTTP MCP server at https://helve.dev/mcp, built-in provider: "fusion" multi-engine search, and $0 billing on failed upstream calls.

Does Helve require Smithery or a third-party MCP wrapper?

No. Helve hosts a native Streamable HTTP MCP server directly at https://helve.dev/mcp with search_tools, call_tool, and request_tool—ready for Claude Code, Cursor, Codex, Gemini CLI, and VS Code with one sk_live_ key.

What happens when an agent needs a tool Helve does not have yet?

Agents can call POST /v1/tools/request or the MCP request_tool endpoint to log missing-tool demand automatically, whereas static gateways require waiting for manual catalog updates.

How does Helve protect agents from runaway spend?

Every sk_live_ key supports a deterministic spend_limit_usd cap enforced synchronously before upstream execution, returning 402 spend_limit_exceeded at the exact threshold, and failed upstream calls debit $0.00.