Your models
Choose the models your agents think with, set fallbacks, bring your own key for a vendor, and see who pays for each call.
Your models is where your workspace chooses the models its agents think with: in chat, at work and on automations. Every plan, Free included, has three ways to pay for a model: Nexa credits, your own key with a vendor, or your own computer through Nexa Bridge. This page helps owners and admins set the order of models, add a key, and understand who pays for each call.
Before you start
- You need to be an owner or admin to change this page. A member reads only Your workspace admins choose the models.
- Open Governance › Settings › Your models.
Understand who pays
- Nexa credits — models served by Nexa, counted against the workspace's monthly credits.
- Your own key — you add a key from a vendor such as OpenAI or Mistral. The vendor bills you directly, and no Nexa credits are counted.
- Your own computer — a model you run yourself, linked through Nexa Bridge. It spends no Nexa credits.
Your own keys and Nexa Bridge share one daily token allowance per workspace, set by the plan.
See what your agents think with now
Open Governance › Settings › Your models.
Read the top of the page: Your agents chat, work and run automations with followed by the models. Under it, a line says who pays, for example Served by Nexa, on your workspace's monthly credits.
Read the three branches of the drawing:
Each shows its model and who pays. Click a branch to open its order below.
Set your workspace's own order
The order your agents try has four choices: Chat, Work, Automations and Agent by agent. Work is open when the page loads. Each of the first three is a chain: a first model and up to two fallbacks. If one doesn't answer, the next one does. Only the one that answers is billed.
Click Chat, Work or Automations.
It reads You're following Nexa's default, or Your chain if you already set one.
Click Choose your own order, or Edit the order if you already have one.
Pick the first model. Search with Search a model or a vendor. Each model says who pays: Nexa credits, Your key or Your bridge.
Click Add a fallback — up to two, tried in order.
Reorder with Try it earlier or Try it later. Remove one with Remove from the chain.
Click Save the order.
The page confirms: Saved. From the next call your agents think with …
To return to Nexa's choice, click Go back to Nexa's default. An empty chain also follows Nexa's default.
A timeout, a vendor error or a busy signal moves on to the next link. Any other refusal stops, and says why.
Give one agent its own order
Click Agent by agent.
Click Open Agent Lab, open the agent and choose its models on its Model card.
An agent's own order wins over the workspace's order. See Agent Lab.
Add your own key for a vendor
In Bring your own key, find the card of the vendor. Each card says where its models are offered, for example Nexa credits · Available and Direct API · Available, and where the vendor runs them.
Click Add your key.
A sheet opens with three steps: Get a key, Paste and save and Done.
In Get a key, use the vendor's links to get the key and read its prices. Click I have my key.
In Paste and save, paste the key. Optionally, type a Label to recognize it later; only you see it.
Keep the Vendor's default address, or click Use a custom address. A custom address must start with https and use a host name, never a bare IP address.
Click Save the key.
Nothing is sent to the vendor now; the first answer on the key tells whether it works.
At Done, those models answer on your key and spend no credits. Click Set the order to place one of them in a chain.
Replace or remove a key
On the vendor card, click Manage.
Click Replace key, or Remove key and confirm with Remove. Keep it cancels.
When you remove a key, its models go back to Nexa credits from the next call, where Nexa serves that vendor. A link of a chain set on your key shows Channel not set up until you add the key again or switch the link to Nexa credits.
Your own models
Your own models · BYOM lists the models you run through Nexa Bridge, and the bridge models Nexa prepares, marked By Nexa · Your bridge. Each shows its place in your chains, or Not in your chain. Without any, it reads No model of your own yet. with Open Nexa Bridge.
A bridge preset sets how Agora runs the model your computer serves. Qwen, Qwen High and Qwen Extra High run the same model and differ in how long it thinks before it answers: the higher the preset, the slower and more careful the work.
Choose models for Design Studio and work without an agent
Open Customize services.
Find the service:
Each service reads where its model comes from: Nexa's default, The agent's model (the model of the agent that makes it) or Your choice.
Click Customize, pick its order and save. Back to default returns it to its default.
What you will see
Good to know
- On Nexa credits, a call counts against your monthly credits. On your key, your vendor bills you and Nexa counts nothing.
- When the credits run out, Nexa's models pause until the period renews or you top up. A link on your key or your bridge keeps answering.
- A link on Nexa credits that has run out stops the call. The chain does not move past it to a key or bridge link behind it.
- Your key never meets the credits limit, but it counts in the workspace's daily allowance for your own models, shared with Nexa Bridge.
- Precedence: Nexa's default, then your workspace's chain for that line, then an agent's own order set in Agent Lab. Each wins over the one before.
- Vendors with a key card: OpenAI, Infercom, Anthropic, Google Gemini, Mistral, xAI, OpenRouter, Qwen, DeepSeek, Groq, Cerebras, Kimi, GLM on Z.ai and MiniMax.
- Prices are the vendor's, on the vendor's page.
- If Nexa has not certified any model of a vendor yet, your key is saved and its models appear in your lists as Nexa certifies them, already on your key.