Skip to content
Bluey
PTENES
Sign in Talk to us
Multi-LLM layer

Each agent runs the modelits own work calls for

You pick the language model for every agent across OpenAI, Anthropic, Google, DeepSeek and xAI, and when you simply want to ask, Synapse picks the best model for you.

Explore a demo See how it works
  • You pick the model, agent by agent
  • Changing the model does not rebuild the agent
  • Usage shows up broken out by model

One model for everything, or the right one for each thing

Drag the line and see the difference between living with a single choice and having the choice on every agent.

The model choice travels with the agent

You set instructions, knowledge, tools and model in one place, and that configuration comes along every time the agent is called, in the chat, inside a squad or in the middle of a workflow.

Agents & Squads

Pick the engine that fits the role

In the agent panel for intelligence and behavior you open the catalog grouped by provider and pick the model, and in the same place you set how much the answers may vary, from precise to exploratory.

  • Models from OpenAI, Anthropic, Google, DeepSeek and xAI in one catalog
  • Search by model or provider name, with a filter per provider
  • Changing the model changes one field, and instructions, knowledge and tools stay as they were
Chat

In the chat, Auto decides when you would rather not

The message box opens on Auto and Synapse picks the best model for that request, and the moment you call an agent with the at sign the picker steps aside, because what answers is the model that agent carries.

  • Auto by default, with Synapse picking the best model
  • You can pin a specific model to the message whenever you want
  • Call the agent with the at sign and its model takes over
Agents & Squads

One squad, several models working together

A squad carries a default model for the agents it creates itself, and every specialist keeps its own, so the same piece of work can pass through different models without anyone having to think about it in the moment.

  • A default model on the squad, so you can start without deciding
  • Every specialist keeps the model you chose for it
  • The squad runs in sequence or in parallel, with a leader closing the synthesis
Analytics

The comparison comes from your use, not from a promise

In Analytics, the costs tab breaks tokens out by model and shows the average input and output per query, which is where the cost concentrates, so comparing one model with another becomes a reading of what your operation has already done.

  • Tokens by model, over the period you choose
  • Input and output separated, to see where the cost concentrates
  • Who gets to read those numbers depends on the workspace permission

How that choice lands in your day

  1. 1

    Describe what the agent does

    The goal, the rules, the boundaries and the response format, in the language you speak.

  2. 2

    Pick its model

    In the same panel, with the catalog grouped by provider and a search by name.

  3. 3

    Use it, compare it, change it

    Usage shows up broken out by model, and changing it is a field, not a fresh start.

We can show this in your own context

Bring a piece of work your team already does today, and we will build the agent, choose the model alongside you and run the conversation in front of you.

Explore a demo

The conversation starts from your case, not from a slide deck.

Changing the model does not change your rules

  • The same safety instructionsEvery call goes out inside the same envelope of rules, whichever model you chose.
  • No AI trainingWhat you send is not used to train AI models.
  • Answers with the sourceWhen the agent answers from the Knowledge Base, the answer shows where it came from.
  • Every run on the recordThe call is audited and accounted for in your workspace, model by model.

On TVBluey, the company behind Synapse, on TECH SC by Globo NSC

Questions about choosing the model

Who picks the model for each agent?

You do, inside the agent itself. The intelligence and behavior panel opens the catalog grouped by provider, and the choice is saved there alongside the instructions and the tools.

And when I simply open the chat?

The message box opens on Auto and Synapse picks the best model for the request. If you would rather decide, you pick a specific model right there before sending.

Can I change the model later?

You can, as often as you like. Instructions, knowledge and tools stay as they were, because the model is one field of the configuration and not the whole agent.

How do I compare two models?

Through use. In Analytics, the costs tab breaks tokens out by model and shows the average input and output per query, so the comparison comes from what your operation has already run.

Do I have to bring my own API keys?

No. The providers come configured with the platform, and the AI usage is accounted for in your workspace.

What happens when a new model generation arrives?

The catalog belongs to the platform and Bluey keeps it current. Your agents carry on with the configuration they have, and you decide when to move.

Pick the model, or let Auto pick it

Either way the decision is yours and it stays recorded on the agent. Let us build this around a piece of work your team already does.

Explore a demo