A single OpenRouter key lets you run hundreds of AI models from many providers in Google Sheets through one credential and one balance, without holding a separate key for OpenAI, Anthropic, Google, and Mistral. In ReplyLabs you paste that one OpenRouter key into the add-on, pick any model OpenRouter fronts, and AI runs across your spreadsheet rows at the provider's published rate. Because this is BYOK (bring your own key), there is no ReplyLabs markup on the model call: the spend lands on your own OpenRouter balance. This article explains how OpenRouter works, why one key is simpler than five, the cost picture, and exactly how to wire it into Sheets.
What is OpenRouter and why use it in Sheets?
OpenRouter is a unified API gateway that fronts 300+ models from every major provider, including Anthropic, OpenAI, Google, DeepSeek, Meta, and Mistral, behind one OpenAI-compatible endpoint. You hold one key and one credit balance, and you switch models by changing a single parameter rather than swapping credentials or rewriting integrations.
For a Google Sheets workflow that is exactly the property you want. AI work in Sheets usually means running one prompt down a column of rows, then sometimes comparing models to see which writes the best output for the lowest cost. With direct provider keys, comparing a Claude model against a Gemini model against a cheap open model means provisioning three separate keys and three balances. With OpenRouter, the same comparison is a model-picker change against one key. For the broader pattern of running prompts across a column, see AI in Google Sheets.
How does one key reach so many models?
OpenRouter sits between you and the underlying providers. Your request hits OpenRouter's endpoint, OpenRouter routes it to whichever provider hosts the model you named, the provider runs the inference, and the result comes back through the same endpoint. From your side it is one key, one balance, and one activity log, no matter which underlying model ran.
That routing also gives you fallback. If a chosen model or provider is unavailable, OpenRouter can try an alternative, and you are billed only for the request that actually succeeded. In a long Sheets batch that resilience matters: a transient provider outage does not stall the whole run.
What does it cost to run OpenRouter through ReplyLabs?
Two layers of cost apply, and it helps to keep them separate.
The first layer is OpenRouter itself. OpenRouter passes through each provider's published per-token price without marking up the inference; what the model catalogue shows is what you pay per token. OpenRouter does charge a small fee when you buy credits (around 5.5%, with a minimum), so that credit-purchase fee is the only overhead on top of raw model pricing. Free models are also available on OpenRouter at zero token cost, subject to rate limits, which is useful for testing a prompt before you spend.
The second layer is ReplyLabs. Because an OpenRouter key is a BYOK key, ReplyLabs adds no markup and no per-row base fee on the AI step. Compare that with managed mode, where ReplyLabs bills the provider's raw cost times 1.25 plus $0.0025 per succeeded row. Worked example: a model that costs $0.0010 per row in tokens runs at $0.0010 per row through your OpenRouter balance under BYOK, versus ($0.0010 x 1.25) + $0.0025 = $0.00375 per row managed. Across 100,000 rows that is $100 of OpenRouter spend versus $375 managed. To plug in your own model rate, use the AI cost calculator.
So the total you pay with an OpenRouter BYOK key is: the provider's raw token price, plus OpenRouter's small credit-purchase fee, and nothing from ReplyLabs on the inference. For the full BYOK cost picture across all providers, see bring your own API key.
How to add an OpenRouter key to Google Sheets
Setup takes under a minute.
- In your OpenRouter dashboard, create an API key. It will look like
sk-or-.... If you want a hard ceiling, OpenRouter lets you set a per-key credit cap and alerts. - Top up your OpenRouter credit balance, or pick a free model first to test without spending.
- In Google Sheets, open the sidebar with Extensions, ReplyLabs, Open sidebar, then go to Settings, Connections.
- Paste the
sk-or-...key into the OpenRouter field and save. The key is encrypted immediately and the field stops showing the raw value. - Set up an AI run and pick any model OpenRouter fronts. The cost preview shows the raw provider rate with no markup, confirming BYOK is active.
- Run as normal. The model call bills your OpenRouter balance; ReplyLabs charges nothing for the AI step.
To swap to a different model, just change the model in the run setup. The same OpenRouter key covers all of them, so there is no second key to paste. For help configuring the prompt and run, see the AI prompts help guide.
Is an OpenRouter key safe in ReplyLabs?
An API key can spend money, so storage matters. ReplyLabs encrypts every stored key at rest with AES-256-GCM, an authenticated scheme that both conceals the value and detects tampering. The key is decrypted only at the moment a call is dispatched on your behalf, is never rendered back to the screen after you save it, and is never written to logs or run history. When you rotate or remove it, the stored value is replaced or deleted. You can also cap exposure on the OpenRouter side with a per-key credit limit, so even a worst case is bounded by the cap you set.
When direct provider keys make more sense
OpenRouter is the broadest single-credential option, but it is not the only one. If you already have a direct relationship and a volume discount with one provider, a direct key keeps billing in one place and lets you use that discount. ReplyLabs also accepts direct keys from OpenAI, Anthropic, Google, and Mistral, so you can mix: a direct key for your primary provider, and OpenRouter for breadth when you want to try models you do not have a direct account with.
For most teams that want to compare and switch models freely from a single balance, though, one OpenRouter key is the simplest BYOK choice. Start at the ReplyLabs home page to see the plans.
Common questions
Can one OpenRouter key run multiple models in Google Sheets?
Yes. A single OpenRouter key reaches 300+ models across providers through one credential. In ReplyLabs you switch models in the run setup without pasting a new key, since the same OpenRouter key covers all of them.
Does ReplyLabs mark up OpenRouter usage?
No. An OpenRouter key is a BYOK key, so ReplyLabs adds no markup and no per-row base fee on the AI step. You pay OpenRouter's pass-through token price plus OpenRouter's small credit-purchase fee, and nothing to ReplyLabs for the inference.
What does an OpenRouter API key look like?
OpenRouter keys are in the sk-or-... format. You create one in the OpenRouter dashboard and paste it into the OpenRouter field under Settings, Connections in the ReplyLabs sidebar.
Is BYOK with OpenRouter available on the free plan?
No. BYOK for AI is a Pro and Scale plan feature. New accounts get a $20 credit to try managed mode, and can move to an OpenRouter BYOK key once on a paid plan.
Can I test a prompt without spending money?
Yes. OpenRouter offers free models at zero token cost, subject to rate limits, so you can validate a prompt on a small sample before pointing the run at a paid model.