FerraWhat is an AI agent?
An AI agent combines a model with instructions, conversation context and tools so it can work toward a task instead of only producing an isolated answer.
A model is the intelligence
The model generates text, analyzes an image or proposes a tool call. Different models have different capabilities, prices and providers. Ferra lets you choose the model separately for each agent.
The job gives it direction
An agent’s task is a persistent instruction. “Find, compare and summarize reliable information” is more useful than “be helpful” because it tells the model what a good answer looks like.
The conversation supplies context
Threads keep related messages together. In Ferra, each agent has its own threads, so a study conversation does not become mixed with an email draft.
Tools make action possible
A tool can prepare an external action, such as an email. That does not mean the AI should act silently. Ferra shows the proposal and requires approval before Apple Mail or Messages opens.
A simple example
Researcher has a research task, a selected model and its own threads. You might ask it to compare three products. It summarizes the information and notes uncertainty, but it does not pretend to have checked a source it never received.
Agents are not infallible
AI can misunderstand context or produce a confident but wrong answer. Check important information, avoid unnecessary sensitive data and report responses that violate your expectations.
Try the idea in Ferra
Start with one narrow job. Write what the agent should produce, what it should avoid and how concise it should be. You can change the model without changing the agent’s identity.