This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
What is the key distinction between AI agents and traditional chatbots?
AI agents can only respond to specific user queries with prescripted answers.
Chatbots follow predetermined conversational paths, while agents dynamically plan and execute multi-step workflows to achieve an outcome.
AI agents require step-by-step human guidance for each action they take.
Which capability enables AI agents to interact with external systems like databases and REST APIs?
Reasoning and planning.
Memory and context.
Tool use.
What communication architecture does the GitHub Copilot SDK use to interact with the Copilot CLI?
The SDK communicates with a cloud-hosted API using REST over HTTPS.
The SDK sends JSON-RPC messages to the Copilot CLI running in server mode as a local process.
The SDK embeds the AI model directly and runs inference locally without any external process.
What is the purpose of the SDK's infinite sessions feature?
To allow an unlimited number of concurrent users to share a single session.
To automatically compact context when the context window approaches its limit, preventing token overflow errors.
To enable the agent to register an unlimited number of tools within a session.
How are tools defined in the GitHub Copilot SDK for .NET?
Using XML configuration files that map tool names to API endpoints.
Using AIFunctionFactory.Create from the Microsoft.Extensions.AI package with a lambda function, tool name, and description.
AIFunctionFactory.Create
Microsoft.Extensions.AI
Using decorator attributes on controller classes that automatically expose methods as tools.
What are the five core architectural components that every AI agent system requires?
Input handler, output formatter, cache, scheduler, and logger.
AI reasoning engine, tools, planner and controller, memory and context, and policies and guardrails.
Database, API gateway, message queue, load balancer, and monitoring service.
What does the OnPreToolUse session hook return to control whether a tool call proceeds?
OnPreToolUse
A boolean value indicating true to proceed or false to cancel.
A PermissionDecision value of Allow, Deny, or Ask.
PermissionDecision
Allow
Deny
Ask
An HTTP status code such as 200 for success or 403 for forbidden.
Which three design areas should a developer focus on when building an agent with the Copilot SDK?
Speed, cost, and scalability.
Context, tools, and policies.
UI design, database schema, and network architecture.
What event type signals that the agent finished processing a request, including all tool calls?
AssistantMessageEvent.
AssistantMessageEvent
ToolExecutionCompleteEvent.
ToolExecutionCompleteEvent
SessionIdleEvent.
SessionIdleEvent
What does the SystemMessageMode.Replace setting do in a session configuration?
SystemMessageMode.Replace
It replaces the AI model assigned to the session with a different model.
It replaces the default system message with a custom system prompt.
It replaces all existing tool definitions with a new set of tools.
Which event type should a developer subscribe to when streaming the agent's response token by token in a chat UI?
AssistantMessageDeltaEvent.
AssistantMessageDeltaEvent
What pattern gives a customer-facing agent access to product information and FAQs beyond the AI model's training data?
Fine-tuning the AI model with the company's proprietary data.
Retrieval-Augmented Generation (RAG) using a knowledge base search tool.
Increasing the model's context window size to include all company documents.
Why should back-office agents use tools for mathematical operations rather than relying on the AI model?
Tools are faster than the AI model at performing complex calculations.
Tools produce reliable results, while the AI model might introduce arithmetic errors.
Tools don't count toward the session's token usage limit.
What approach should be used when a back-office agent is triggered by monitoring alerts or webhooks?
Process all alerts in a single shared session for efficiency.
Queue all alerts and process them in a daily batch job.
Process each event in an independent session to prevent context mixing between unrelated operations.
What iterative workflow does an AI agent follow when processing a request?
Research, implement, test, deploy, and monitor.
Receive input, reason and plan, act, observe, and iterate or respond.
Parse request, validate data, execute query, format output, and return response.
You must answer all questions before checking your work.
Was this page helpful?
Need help with this topic?
Want to try using Ask Learn to clarify or guide you through this topic?