> For the complete documentation index, see [llms.txt](https://oodlz-ai-studio.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oodlz-ai-studio.gitbook.io/docs/inbound-agent/the-knowledge.md).

# The Knowledge

A standalone knowledge and tool server — connectable from any MCP-compatible client, with separate retrieval namespaces for FAQ, instructional content, and the live brand catalogue.

> **The agent's knowledge lives in its own server, not bolted into the agent workflow.**

Most AI agents bundle their tools and knowledge directly into the agent. That works until a second agent, a third channel, or a new interface needs access to the same knowledge — at which point the work doubles. The Oodlz inbound system runs the knowledge layer as a standalone MCP server. Any MCP-compatible client connects to it and gets instant access to every retrieval tool without duplicating embeddings, namespaces, or escalation routing.

## How It Works

The knowledge layer exposes 4 tools at one URL: `brands`, `faq`, `other`, and `human_escalate`. Any MCP client — the inbound agent, a future voice agent, a future internal tool — points at that endpoint and the tools are immediately available. Adding a new client means one MCP Client node configured against the existing URL; no rebuilding the vector store, no reconnecting embeddings models, no re-implementing the human handoff.

Underneath, the brand catalogue, FAQ knowledge, and instructional content live in 3 separate retrieval namespaces, each tuned for its query type. Knowledge upload is a separate workflow — anyone with form access can drop a CSV, PDF, or document and have it embedded and live in the agent within seconds.

The pages below cover the layer's parts in detail — the namespace design, the upload pipeline, and the human escalation handoff.

## What You Get

* One knowledge server shared across every current and future channel — no duplicated embeddings, no parallel namespaces.
* Tuned retrieval per query type — brand catalogue, FAQ, and instructional content each have their own settings.
* Form-based knowledge upload — any document format, embedded and live in seconds.
* Structured human escalation as a tool, not a hardcoded fallback — the agent collects context and hands off cleanly.