Admin Guide
What this guide is
The Admin Guide covers every feature on the ESP Private Agents Admin Dashboard — the web interface where admins create, configure, and publish AI agents for their deployment.
If you are responsible for building agents, setting up tools or knowledge bases, configuring the platform, or reviewing usage — this guide is for you.
This guide covers the Admin Dashboard only. If you are an end user who chats with agents, see the End User Guide. If you plan to deploy or operate Private Agents in your own AWS account, please contact Espressif Sales.
What an admin does
An admin builds and manages AI agents on the Admin Dashboard. That means:
- Creating agents — defining an AI assistant with its own name, behaviour, model, tools, and knowledge.
- Choosing a language model (LLM) — picking the model that powers each agent's responses.
- Writing system prompts — the standing instructions that shape how an agent talks and what it focuses on.
- Attaching tools — adding capabilities like web search or MCP servers the agent can call during conversations.
- Setting up Knowledge Bases — uploading documents and URLs the agent can search for answers.
- Publishing agents — sharing agents with end users via a link or QR code.
- Monitoring usage — reviewing costs, requests, and activity through the Analytics page.
Super-admins also manage the deployment itself — configuring LLM provider API keys, managing other admins, and handling stack updates.
Dashboard navigation
The Admin Dashboard has three sections in the navigation drawer, plus a Profile page accessible from the account menu:
| Section | How to reach it | What it is |
|---|---|---|
| Agents | Nav drawer | Your main workspace. Create, edit, configure, and share AI agents. |
| Analytics | Nav drawer | Usage metrics — costs, requests, and activity across every agent and user. |
| Deployment | Nav drawer | Stack updates, LLM provider API keys, admin management, auth settings. Super-admins only. |
| Profile | Top-right avatar button → Profile | Your admin account — name, email, and account details. |
How the agent editor works
Every agent is edited on a single page with two modes, toggled by a pair of icon buttons in the top-right of the editor header:
- Basic mode (form icon) — a structured form with three sections: Basic Info, Tool Integration, and Knowledge Base Integration. Most admins use this.
- Advanced mode (code icon) — a JSON editor showing the full agent configuration. For bulk edits, copying configs, and power-user tweaks.
You switch freely between modes. Your changes carry over.
What's in this guide
This guide is organized to follow the Admin Dashboard flow:
Agents
Everything about creating, configuring, and publishing agents:
- Your AI Agents — the agents list: creating, editing, sharing, and deleting agents.
- LLM (Model) — picking the language model that powers an agent.
- System Prompt — writing the instructions that shape an agent's behaviour.
- Tools — attaching MCP tools and local device tools to an agent.
- Knowledge Base — connecting documents and web URLs the agent can search.
- Advanced Mode — the JSON editor for power users.
- Templates — pre-configured agent and tool templates.
- Sharing — generating share links and QR codes for end users.
Platform
- Analytics — usage costs, requests, and activity across your deployment.
- Profile — your admin account details.
Super-admins who manage the deployment itself (stack updates, LLM provider keys, auth, domains) should use the Deployment Settings guide.
How-to guides
Step-by-step task guides for common workflows:
- Create an agent
- Pick the right LLM
- Add an MCP tool
- Set up a Knowledge Base
- Copy an agent
- Share an agent
Where to start
If you're new to the Admin Dashboard, start here:
- Your AI Agents — understand the agents list and how to create your first agent.
- LLM (Model) — pick the right model.
- System Prompt — write effective instructions.
- Create an agent — a full walkthrough from start to share.