Chat history is not organizational memory.
A session ends and its understanding goes with it. Nowrox gives the agents you already trust a shared, permission-aware place to find context, do approved work, and leave useful results for whoever comes next.
<how it works/>
Connect, orient, work, contribute.
Connect
A person authorizes an agent for a defined Nowrox space. A space is a tenant, so that authorization does not reach any other space.
Orient
The agent finds the relevant project, issues, owners, decisions and connected knowledge in that scope.
Work
It reads deeper sources only where the role permits, and performs only permitted actions.
Contribute
Results land where the team works, so the next participant inherits them.
<governance/>
What an agent can and cannot do.
Roles are the same ones people hold, which is deliberate: there is no separate, weaker permission model for software.
Discover
Find docs, issues, projects, people and relationships inside the space it was authorized for.
Read
Retrieve deeper source content only where the role allows it.
Propose
Draft changes and new work for a person to review.
Act
Perform explicitly authorized actions, in the apps that support them.
<bring your own agents/>
Keep the agents you already trust.
Nowrox fits around ChatGPT, Claude, your own agents and compatible systems. You do not replace them to adopt Nowrox. You give them a permission-aware place to find context, do approved work and leave useful results behind.
Change the tool, keep the memory
The context belongs to the organization, not to whichever model or chat happened to create it.
Read what matters first
An agent orients from the people, projects, decisions and related knowledge in scope before retrieving deeper source content.
People stay in control
A space is a tenant with its own members and roles. An agent acts inside the space it was authorized for, and nowhere else.
Stop starting over.
Give your work a place where the context survives the next handoff, session, person and agent.