What is Qoni?
Qoni is a documentation surface for agent infrastructure. It does not describe one agent application. It explains the underlying systems that give agents identity, memory, and action in production.
One-line definition
Qoni's core narrative is simple: Build Agent infrastructure.
It breaks production agent capability into three layers: identity boundaries, persistent context, and real-world execution. The docs site exists to explain how those layers work together.
Capability matrix
| Layer | Product | Responsibility |
|---|---|---|
| Identity | GenAuth | Human-to-Agent authorization, MCP Profile, audit, and compliance |
| Action | Web Agent | Search, extraction, Textify, browser actions, and sandboxed execution |
| Memory | GUMem | Chat Memory, Profile Memory, Action Memory, and Search Memory |
Why these layers belong together
Traditional application infrastructure is designed around human users: login, permission, database, API gateway, and monitoring. Agent applications need a different set of assumptions:
- Agents act on behalf of users, so authorization must be delegated, scoped, revocable, and auditable.
- Agents run long tasks across sessions, so context cannot live only inside prompts.
- Agents need real web access and third-party tools, so action must become engineered infrastructure.
Qoni turns these needs into a clear pipeline: intent enters the system, identity defines boundaries, memory provides context, and action connects the external world.
What you will find in this docs site
- The homepage builds the shared model for the stack.
- Module pages explain the responsibility and boundary of each layer.
- Overview pages describe architecture, scenarios, and system shape.
- Research pages capture adjacent market context, benchmarks, and directional judgment.
Product posture
Qoni should read like infrastructure: restrained, credible, production-oriented, and composable. Pages should explain boundaries, operating constraints, and next steps before they try to sound impressive.