AfriGlade Documentation

Getting Started

AfriGlade gives you one place to chat with many large language models and deploy ready-made AI Agents.

1. Go to chat.afriglade.com
2. Create an account or sign in
3. Choose a model and start chatting, or pick an Agent from the marketplace

Using Chat

Pick any available model from the model picker. Your history is saved per conversation. Routers such as smart-router choose a suitable model for you.

• Be specific about context (for example “for a Nigerian SME” or “in Naira”)
• Use an Agent when you want a ready-made role and instructions
• Break complex tasks into smaller steps

Agents

Agents are custom AI assistants with their own instructions, model and tools. They work like GPTs or Assistants, but can run on many model providers.

1. Browse the Agents Marketplace
2. Click “Deploy” on the Agent you want
3. In chat, you can also type @ and the Agent name to call it

Building your own Agent

Choose “Agents” from the endpoint menu and open the Agent Builder in the side panel. The form includes:

• Avatar, Name and Description to identify the Agent
• Instructions that define its behaviour, typed inline or linked from a saved Prompt
• Model and parameters such as temperature and maximum output tokens
• Conversation starters, up to four suggested prompts shown on a new chat
• Tools and Skills you pick from a searchable library

Capabilities

• Code Interpreter runs code (Python, JavaScript, TypeScript, Go, C, C++, Java, PHP, Rust and Fortran) and handles file uploads and downloads
• File Search answers from your uploaded documents using semantic search
• File Context stores text extracted from files in the Agent’s instructions

Some capabilities, such as stateful code sessions, are experimental and may not be enabled on every deployment. Users with edit access can also use Version History to review or restore earlier Agent configurations.

MCP Tools

The Model Context Protocol (MCP) is an open standard for connecting AI models to outside tools, data and services, a bit like a universal plug for AI. An MCP server can give an Agent things like file access, web browsing or your business APIs.

In a normal chat

Select a tool-compatible model, then choose servers from the MCP Servers section of the + menu in the message box. Each selected server shows as a chip, and you can use several at once.

With an Agent

1. Open the Agent Builder and choose Add tools, then MCP
2. Select a server (each server is one entry)
3. Expand it to switch individual tools on or off
4. Save the Agent

Adding a server from the interface

1. Open the MCP Settings panel in the right sidebar and press +
2. Enter the name, description, URL, transport type and authentication method, then Create
3. If the server uses OAuth it shows as disconnected. Select it in the + menu, then choose Continue with OAuth and complete the provider’s sign-in
4. When authentication succeeds the server is connected and selected for your chat

Agents with many MCP tools can mark tools as deferred, so they are only loaded when the model searches for them. This keeps the context window free.

Scheduled Chats

Scheduled Chats let a saved Agent start a new conversation automatically with a prompt you write, for recurring reports, reminders or research. This is an experimental feature, so it only appears if it is enabled on your account.

• Each schedule needs a name, a prompt, a saved Agent and a time zone
• Cadence can be hourly, daily, weekdays, weekly or a custom five-field cron expression
• You can edit, pause, run now or delete your schedules, and each run opens as its own conversation
• A run can be skipped if another run of the same schedule is active or your balance is too low
• Repeated failures disable a schedule automatically until you fix the cause
• Scheduled runs use tokens like any other chat, so they are billed the same way

Pricing & Billing

AfriGlade is pure pay-as-you-go, based on input and output tokens. Prices are shown in Nigerian Naira (₦) at ₦1,550 per $1.

See the Pricing page for current rates per model and a cost estimator.

Usage Statistics

The Pricing page shows platform-wide token usage and cost for the last seven days, and your account shows your own usage so you can stay in control of spending.

Frequently Asked Questions