AgentsAgents are the core building block of Prospector Studio. Each agent combines an LLM with custom tools, skills, and knowledge bases to perform tasks autonomously.
Creating an Agent
From the Studio UI, navigate to the Agents section and create a new agent. Configure:
- Name — A descriptive name for the agent
- Model — The LLM provider and model (e.g., Claude)
- System Prompt — Instructions that define the agent's behavior
- Tools — External integrations the agent can invoke
- Knowledge Bases — Document collections the agent can query for context
Agent Capabilities
Tools
Agents can execute tools to interact with external systems. Tools are defined as actions the agent can invoke during a conversation, such as querying a database, calling an API, or triggering a workflow.
Knowledge Base Integration
Attach one or more knowledge bases to an agent to give it access to your organization's documents. When a user asks a question, the agent performs a semantic search against the knowledge base and includes relevant context in its response.
Conversations
Each conversation with an agent maintains full context history. Agents can handle multi-turn interactions, reference previous messages, and maintain state across a session.
LLM Providers
Prospector Studio reaches models through AWS Bedrock, so the catalog spans every provider Bedrock carries — Anthropic, Amazon, Meta, Mistral, Cohere, DeepSeek, Qwen, and more — behind one unified API. Alongside those, privately fine-tuned models can be hosted on your own network.
The provider and model are configured per agent, so different agents can use different models based on their use case, and you can change an agent's model without redeploying it.