Agents
Agents that act, not just answer
Each BrainChat agent reads your operation's tables, fills in data, calls integrations and knows when to hand over to a person.
AI with data and actions
Flow-based chatbots get stuck when the conversation leaves the script. The BrainChat agent uses Google Gemini with knowledge base search (RAG) and configurable actions (function calling): it looks up orders, creates records, schedules, updates stages and triggers integrations.
The AI runs on the customer's own API key. There is no per-conversation AI charge, and the company controls the model and the cost.
Features
- No-code actions: Close a collection, schedule, create a card, update a record, call an external API.
- Knowledge base: Documents and Wiki pages indexed for answers with sources.
- Data extraction: The agent fills in table fields from the conversation.
- Test lab: Test the agent before publishing, with versions.
- Handoff to a person: By rule, by intent or at the customer's request, with context.
- Own language: Each agent can answer in Portuguese, English or Spanish.
Frequently asked questions
Which AI model do the agents use?
Google Gemini, with the customer's own API key.
Can the AI agent update the CRM?
Yes. It reads and writes BrainChat tables and calls external APIs through the configured actions.
When does the agent hand over to a person?
When the configured rule says so, when it detects the intent or when the customer asks; the conversation stays in the same inbox.