Overview
Hopscotch is an LLM gateway that collapses the multi-provider integration tax into a single OpenAI-compatible endpoint. Instead of signing separate contracts, wiring distinct SDKs, and reconciling five dashboards, a developer points their existing OpenAI client at the Hopscotch base URL, swaps the model name, and suddenly has 500+ models from Anthropic, OpenAI, Google, xAI, DeepSeek, Perplexity and more behind one API key and one prepaid balance. What makes it more than a reseller is the cost-control layer: spend limits per key, per teammate, and per workspace that reset on a schedule and reject over-limit requests before they ever reach a provider. In our evaluation this is the feature that matters most for production teams running autonomous agents, where a single runaway loop can otherwise torch a budget. Hopscotch states it never swaps the model you request, never edits your prompts, and charges provider list prices with no markup, so the savings come from governance rather than from squeezing margin out of tokens.
Key Features
- One OpenAI-compatible base URL, one API key and one balance for 500+ models across major providers
- Routing profiles with automatic fallbacks when a primary model errors or rate-limits
- Per-key, per-teammate and per-workspace spend limits that reset daily, weekly or monthly
- Activity log recording every request with serving provider, tokens, cost and latency, exportable to CSV
- Side-by-side model lab to compare answers and cost on your own prompts before changing code
- Bring-your-own provider keys for free, billed directly by the upstream provider
Pricing
| Plan | Price | For |
|---|---|---|
| Pay-as-you-go | Credit from $5 | Buy prepaid credit, pay each provider’s list price, no subscription |
| Bring your own keys | $0 | Use your own provider keys free, provider bills you directly |
| Enterprise | Custom | High-volume governance, SLAs, dedicated support |
Comparison
Against a model marketplace like OpenRouter, Hopscotch is deliberately minimal: the same no-markup philosophy, but the emphasis is spend governance and failover for agent workloads rather than community model discovery. Compared with managed inference platforms such as Together or fal, Hopscotch does not own GPUs; it is a routing and billing layer that lets you mix those providers plus many others through one integration.