Add a model key
Add an OpenAI or OpenRouter key so tracked prompts can run, where account keys and project keys differ, and what the runs cost.
Updated October 20, 2026
Omnipresence runs on your key, not ours. You pay the model provider directly at list price, and we never resell or mark up a model call. That is why tracked prompts are unlimited on every plan.
Which key covers what
| Key | Assistants it covers |
|---|---|
OpenAI (sk-...) | ChatGPT |
OpenRouter (sk-or-...) | Claude, Gemini, Perplexity |
Two keys, four assistants. One key gets you a partial picture, which is fine for a first look and misleading as a baseline: the four disagree with each other more than most teams expect, and the one you skip is often the one your buyers use.
Add it
1. Get a key from the provider
Sign in to OpenRouter or OpenAI and create an API key. Both show the key once, so copy it before closing the dialog. Fund the account with a small amount; both bill against a balance and a key with no credit fails every run.
2. Decide account or project
Account keys live under Account and apply to every project you own. Project keys override the account key for one project.
One key for everything is right for an in-house team. A key per project is what agencies want, because the run cost then lands on the client's own provider bill rather than being something you absorb and re-invoice. Per-project keys are an Agency feature.
3. Paste it in
Open Account, find Model API keys, paste the key into the OpenAI or OpenRouter field and save. The key is validated on save: a typo is rejected there rather than silently failing on the next scheduled run.
4. Check what a run will cost
The prompt dialog shows an estimated monthly cost as you set the cadence and the sample count, across all platforms. Read it before you set forty prompts to daily.
As a rough shape: a prompt costs a few cents per assistant per run. Twenty prompts, four assistants, weekly, is a couple of dollars a month. The same twenty on daily is seven times that. Cadence is the dial that moves your bill.
Changing a cadence changes your model spend the same day rather than at the start of the next month. Nothing stops you putting two hundred prompts on daily; your provider will bill you for it.
If a run fails
A failed run is almost always the key: no credit, a revoked token, or a rate limit on the provider side. Transient rate limits are retried automatically. A persistent failure shows on the project with the provider's own reason, which is the fastest thing to read before you suspect anything else.
Questions about add a model key
- Why bring my own key at all?
- Because every prompt costs money to run. A tool that bundles that cost has to guess high and cap your prompts. We charge for the software and let you buy model calls at list price, which is cheaper and visible.
- Do I need both keys?
- Only if you want all four assistants. OpenAI covers ChatGPT; OpenRouter covers Claude, Gemini and Perplexity through one key.
- Is the key safe?
- Keys are encrypted at rest, used only to run your own scheduled prompts, and never logged or sent anywhere else. You can replace or remove one at any time.