A LiteLLM alternative with nothing to host
The LiteLLM proxy is a self-hosted server you run yourself, deployed with Docker, Helm or Terraform. Hopscotch is the managed option: one base URL, 0% fee, and a budget on every key.
LiteLLM routes to providers on your own keys, so you pay each provider directly and pay for hosting. License and Postgres per LiteLLM docs, Nov 6, 2026.
0% fee on every top-up.
Credit from $5, and every token at the provider's list price. Or pick a Pass: each seat gets twice its price in model use every month.
A budget, fallbacks and guardrails on every key.
Refused and blocked requests cost $0.
A 99.9% availability target, with service credits.
On every account, not only an enterprise tier.
- Anthropic
- OpenAI
- DeepSeek
- Qwen
- Meta
- Mistral AI
- Moonshot AI
- MiniMax
- Z.ai
- Cohere
Hopscotch vs LiteLLM
Hopscotch is an LLM gateway: one API key for 500+ models from Anthropic, OpenAI, Google and DeepSeek, with spend limits, fallbacks and request logs built in. It is a LiteLLM alternative for teams that want the single endpoint without a proxy to host.
Budgets on every key, with no proxy to run
Every key, teammate and workspace gets its own cap. A request over it is refused before it reaches the provider, so it costs $0. With LiteLLM, budgets live in a proxy and Postgres database you host and keep patched.
Passes: 2x what you pay in model use
Each seat gets twice its price in model use, on every model, frontier included, at provider rates with no markup. LiteLLM passes your provider's price straight through, so $50 buys $50 of model use, plus whatever it costs to host the proxy.
Monthly, per seat.
Hopscotch vs LiteLLM: where the routes split
| Feature | Hopscotch | LiteLLM |
|---|---|---|
| Where it runs | Managed by Hopscotch | Your servers; an enterprise offering also exists (LiteLLM docs, Nov 6, 2026) |
| Setup | Change a base URL and key | Deploy the proxy, then add provider keys; the docs require a Postgres database for keys and budgets |
| Provider accounts | One Hopscotch balance, or your own keys | Your own key at each provider |
| Spend controls | Budget on every key, refused requests cost $0 | Budgets per key, team, org and model (LiteLLM home page, Nov 6, 2026) |
| Fallbacks | Up to five models per routing profile, tried in the order you set, changed in the dashboard without a deploy | Fallbacks and load balancing (LiteLLM home page, Nov 6, 2026) |
| Uptime commitment | 99.9% monthly availability target with service credits, on every account | Depends on how you host it |
Three questions that decide it
Hopscotch is managed, so there is no server, database or upgrade to run. It works with any OpenAI-compatible tool, and the uptime target is ours to keep.
See setup guidesEvery Hopscotch key carries its own budget, and a request over the limit is refused before it reaches the provider. The loop stops, and the refused requests cost nothing. Teammates and the whole workspace get ceilings too.
See spend controlsOn Hopscotch, requests on your own OpenAI, Anthropic or Google key are billed by that provider, with no Hopscotch fee at any volume.
See bring your own keyHow to switch from LiteLLM in four steps
- 01
Sign up, add credit from $5, and copy the key from the dashboard.
- 02
In the OpenAI SDK you already use, set base_url to https://api.hopscotchlabs.ai/v1 and api_key to your Hopscotch key. Nothing else in your client changes.
- 03
Model IDs take the form provider/model, such as anthropic/claude-sonnet-5. A bare ID is refused with a 400 so nothing silently routes to a provider you did not pick. Pull the list from GET /v1/models or the models catalog.
- 04
Rebuild your LiteLLM model list and virtual-key budgets as Hopscotch routing profiles and key budgets. Put backups in a routing profile and call it as profile/your-name. Set a routing profile as your default and it calls that profile in the order you set; until you do, it answers from Hopscotch's balanced shortlist for your workspace.
Your coding agent moves the same way. Setup guides exist for OpenCode, Cline, Codex CLI, Cursor, Aider, Continue and Zed. The quickstart walks through your first request.
from openai import OpenAI
client = OpenAI(
base_url="https://api.hopscotchlabs.ai/v1", # was: your LiteLLM proxy URL
api_key="YOUR_HOPSCOTCH_KEY", # was: your LiteLLM key
)
# model ids gain a provider: anthropic/claude-sonnet-5
client.chat.completions.create(model="anthropic/claude-sonnet-5", messages=[...])How Hopscotch compares with other LiteLLM alternatives
Here is what each gateway costs you, in fees or in work.
More comparisons: OpenRouter alternative, Portkey alternative, Vercel AI Gateway alternative, OrcaRouter alternative and Cheaper Inference alternative.
| Gateway | What it is | What you take on |
|---|---|---|
| Hopscotch | Managed gateway: one key, provider list price, budgets and guardrails on every key | Nothing on top: 0% fee on credit |
| LiteLLM | Open-source proxy on your own provider keys | Hosting, scaling and patching the proxy yourself |
| OpenRouter | Managed gateway with public usage rankings and a large app ecosystem | A 5.5% to 8% fee on every top-up |
| Vercel AI Gateway | Managed gateway that works from any server, with no markup or platform fee on tokens | A Vercel team account, and card processing fees where they apply |
| Prisma AIRS AI Gateway (formerly Portkey) | Gateway built around enterprise governance and observability | A governance platform to set up before the first request |
| Cloudflare AI Gateway | Proxy for caching, logging and rate limits | A 5% fee on Cloudflare credit, or a key and bill per provider |
| Direct provider APIs | Each provider's own API and key | No failover, and a key and bill per provider |
LiteLLM alternatives: questions people ask before switching
Is Hopscotch a LiteLLM alternative I do not have to host?
Yes. Hopscotch is managed, with one base URL, a 0% fee on credit and no server to run. LiteLLM is self-hosted.
Does Hopscotch have monthly plans?
Yes, and they are optional. Hopscotch Passes are per-seat monthly plans where each seat gets twice its price in model use, on every model, frontier included, at provider rates with no markup: Starter is $20 a month for $40 of model use, Growth $50 for $100 and Pro $200 for $400. Limited time offer. Credit from $5 still works with no pass.
How does LiteLLM pricing compare?
Hopscotch has no hosting and no fee on credit. The LiteLLM proxy is open source under an MIT license, so you pay for hosting and provider usage.
What does LiteLLM do that Hopscotch does too?
Both give one OpenAI-compatible endpoint with budgets, fallbacks and guardrails. Hopscotch runs it for you. LiteLLM's home page lists PII masking and prompt-injection guardrails (Nov 6, 2026).
Does LiteLLM have an enterprise version?
Hopscotch includes budgets and guardrails on every account. LiteLLM's home page lists an Enterprise tier that adds SSO, JWT auth, audit logs and custom SLAs, with pricing on request (Nov 6, 2026).
Is Hopscotch's source code open?
No, Hopscotch is closed source. LiteLLM is open source under an MIT license.
LiteLLM vs OpenRouter: where does Hopscotch fit?
Hopscotch is managed like OpenRouter, but charges 0% on credit instead of a fee on top-ups. OpenRouter's fee is 5.5% to 8%, read Nov 6, 2026.
When is LiteLLM the better fit?
Hopscotch is managed only. LiteLLM fits when you must run the gateway inside your own network or change its source.
Can I move from LiteLLM to Hopscotch?
Yes. Set the base URL to https://api.hopscotchlabs.ai/v1, paste your key, and use provider/model names.
Same models, one base URL change, nothing to host
Connect your SDK to Hopscotch, add credit and send a request. If running the proxy was the reason you looked, there is nothing left to host.