The Agentic Builders / New to Agents · 1 of 14
What is an Agent
· 3 min read
New to Agents · Beginner Path · 1 of 14
People say "AI agent" like it names one thing. It doesn't. An agent is an assembly - and once you can see the parts, the whole field stops being mysterious.
Start with the difference that matters. A chatbot answers a question and stops. An agent does work: you hand it a goal, and it breaks the goal into steps, takes actions, checks the results, and keeps going until the job is done. The chatbot is something you talk to. The agent is something you delegate to.
What makes that possible is not one magic component. It's four parts working together:
- An LLM - the reasoning engine. The part that reads the situation and decides what to do next.
- A harness - the program wrapped around the model that lets it actually act: read and write files, run commands, call tools, and loop instead of answering once.
- Tools - the specific capabilities it can reach for: search the web, edit a file, query data, send a message.
- Memory - so it carries context forward instead of starting from zero every time.
Snap those together and you get something you can hand a real job: not a clever conversation, but a worker.
Why this matters for Tropo
The first two parts - the LLM and the harness - come from the labs. Anthropic, and the tools built on its models, give you the engine and the hands.
The last two are where Tropo lives. Tropo is the layer that gives an agent a memory that survives (durable files in a navigable vault), work it can be held accountable to, and a crew it can be part of. The lab gives you a capable worker for one session. Tropo is what turns it into a teammate that remembers, improves, and coordinates across thousands of them.
Beginner Path · New to Agents Next: What is an LLM (the engine at the centre) → Related: What is Memory · What is a Harness · See the whole path: New to Agents - Beginner Path