> ## Documentation Index > Fetch the complete documentation index at: https://help.iteam.works/llms.txt > Use this file to discover all available pages before exploring further. # Project Agents > How agent access works inside a project: the project agent and the agents added to it can see the project's knowledge, objectives and data — those outside cannot. Isolation explained simply. Inside a project, agents only see **what belongs to the project**. This page explains, in simple terms, **who accesses what**: the project agent and the agents you add to it have access to the project's context; agents that are **outside** the project don't — even if they're powerful elsewhere in the company. A project's agents ## The project agent Every project can have a **project agent** — the specialist for that context. It's the one that **measures the KPIs**, answers in the project chat and works the objectives, always with access to the knowledge and data of **that** project. The project agent only sees the project's world. It doesn't pull knowledge from the company or from other projects — unless you've explicitly linked something (for example, an objective with **"Derives from"**). ## Who sees the project's context The project agent and the agents/teams you **added** to the project can access the project's **knowledge**, **objectives**, **actions**, **results** and **data**. Any agent that is **not** part of the project **doesn't** see any of it. For it, the project's content simply doesn't exist. ## How to add an agent to the project In the project, use the **agents and teams** panel on the right. Choose from the agents/teams you have access to. From then on, it becomes part of the project and sees its context. Click the agent to open the chat already in the project's context. It answers based on the project's knowledge and objectives. ## Isolation, explained simply Think of each project as a **closed room**. Inside the room are that project's documents, goals and data. Only those **in the room** (the project agent and the agents you put there) see what's inside. Those outside see nothing — and an agent that's in two different rooms **never** carries content from one to the other. Isolation isn't just "hiding it on screen". Knowledge search and KPI measurement are **scoped to the project** in the mechanism itself: an agent from outside can't bring, in an answer, content from a project it isn't part of. ## Isolation between projects The same agent can take part in **several projects**. In each one, it only sees **that** project's context. It never crosses information — what it learned or read in Project A doesn't leak into Project B. ## People's permissions Agent access goes hand in hand with **people's** permissions: * Whoever has access to the **project** sees the project's agents and content (respecting the access to each agent). * Company **Owner/Admin** see the projects according to the company's access rules. * Granting someone access to the project is a practical way to release, all at once, the resources tied to it. See [Security & Access](/en/administration/security-and-access). ## Next steps The knowledge and objectives the project's agents see. Control who accesses what in the company and in projects.