Model Keys and Dedicated Capacity
How Build For You and Fractional CTO organizations get a dedicated model capacity lane and can bring their own Google, OpenAI, and Anthropic keys with no platform fee.
On the Build For You and Fractional CTO plans, your agents get two model benefits:
- A dedicated model capacity lane. Your traffic runs on capacity that is separate from the shared pool, so a busy day for other customers does not slow you down.
- Bring your own keys, with no platform fee. Connect your own Google, OpenAI, and Anthropic keys. Your agents run on your own quota and your own provider agreements, and the provider bills you directly.
Builder, Studio, and Studio Pro run on Chipp’s shared capacity and cannot connect their own model keys.
Dedicated capacity lane
Every model provider limits how many tokens one account can send per minute. On the shared pool, a large burst of traffic from any customer counts against the same limit.
On Build For You and Fractional CTO, Gemini models run on a dedicated Google Vertex AI lane, separate from the shared pool. There is nothing to set up: the lane turns on with your plan, and Settings → Model Keys shows it as active.
Bring your own keys
Which keys you can connect
| Provider | Key type | Where to get it |
|---|---|---|
| Google AI Studio API key | Google AI Studio | |
| OpenAI | OpenAI API key | OpenAI platform dashboard |
| Anthropic | Anthropic API key | Anthropic Console |
Connect a key
- Go to Settings → Model Keys. Only organization owners and admins can add, rotate, or remove keys. Other members see a read-only view.
- Paste the key for a provider and save. Chipp checks the key with one quick call to the provider before it stores it. If the check fails, nothing is saved.
- Choose what happens if the key fails (see below).
After you save a key, Chipp never shows it again. The page shows only the last four characters, the key status, and when Chipp last checked it. Keys are encrypted at rest. To change a key, rotate it. To stop using it, disconnect it.
What runs on your key
When you connect a key for a provider, your agents’ chat responses on that provider’s models run on your key.
Background work still runs on Chipp’s own keys and bills as usual. Examples: conversation titles, memory, knowledge source embeddings, tools, and image generation.
What it costs
Chipp adds no platform fee on model usage that runs on your own key. Your provider bills you for those tokens at your own rates. Chipp does not deduct them from your credits.
Usage that runs on Chipp’s capacity, including the background work above, bills at standard rates.
If your key fails
For each key you choose a fallback:
- Use Chipp’s capacity (default). If your key is rejected, rate limited, or the provider has an error before the response starts, Chipp retries the turn once on its own capacity, on your dedicated lane. That turn bills at standard rates.
- No fallback. The response fails instead of retrying on Chipp’s capacity.
If the provider rejects the key itself (for example, it was revoked), Chipp marks the key as invalid and shows the error on the Model Keys page. Connect a new key to fix it.