You can create an agent from the Agents page in your dashboard. Click Creator Wizard to generate an agent from your content sources, or Classic Setup to configure each option yourself.Documentation Index
Fetch the complete documentation index at: https://docs.brainstormer.io/llms.txt
Use this file to discover all available pages before exploring further.
- Classic Setup
- Creator Wizard
Classic Setup walks you through six steps. You can move forward and backward freely — earlier steps are saved as you go.
Basic info
Enter a name for your agent (required) and an optional description.Set the status to Draft if you want to configure and test the agent before making it live, or Published to activate it immediately.Agent names must be 255 characters or fewer. Descriptions can be up to 2,000 characters.
Select model
Choose the AI model that powers your agent. Brainstormer gives you access to 300+ models from providers including OpenAI, Anthropic, Google, Meta, Mistral, and more.Use the provider filter to narrow the list. Each model card shows context window size, input/output pricing, and recommended use cases.Model selection tips:
- For fast, high-volume use cases (customer support, FAQ bots), smaller or free-tier models reduce cost with minimal quality loss.
- For complex reasoning, document analysis, or multi-step tasks, larger models like GPT-4o or Claude 3.5 Sonnet perform better.
- If you attach a knowledge base, choose a model with a larger context window so it can incorporate more retrieved content.
System prompt
The system prompt is the set of instructions your agent follows in every conversation. It defines the agent’s persona, scope, tone, and any rules for how it should respond.Template variables. The prompt editor supports Mustache syntax for dynamic values. For example, The editor shows a character count and warns you when the prompt is approaching the model’s context limit.
{{user_name}} is replaced at runtime with the actual user’s name when you pass it through the API or conversation variables.Voice setup (optional)
Clone a custom text-to-speech voice for your agent, or skip this step and use the default TTS voice.Click Create Custom Voice to open the voice upload interface, or Skip Voice Setup to continue without one. You can add or change a voice at any time from the agent’s edit page.See the voice cloning guide for full details on uploading audio samples and configuring voice settings.
Knowledge bases (optional)
Link one or more knowledge bases to your agent. When linked, the agent automatically searches the knowledge base on every message and includes relevant content in its response.Toggle any knowledge base from your organization to link or unlink it. You can attach multiple knowledge bases if your content spans different collections.
Knowledge bases must be created separately before you can link them here. Go to the Knowledge section to create one.
Review and save
The preview step shows a summary of your agent’s configuration: name, model, system prompt, voice setup, and linked knowledge bases.You can also open a live chat window to test the agent before publishing. This lets you verify the system prompt and knowledge base responses before your agent is live.Click Save Agent to create the agent. You are taken back to the Agents page where your new agent appears in the list.