You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/copilot/customization/language-models.md
+5Lines changed: 5 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -25,6 +25,11 @@ Use the language model picker in the chat input field to change the model that i
25
25
26
26
27
27
28
+
> [!TIP]
29
+
> AI Toolkit can expose language models to enhance GitHub Copilot capabilities.
30
+
>
31
+
> For more information, see [Change the chat model](https://docs.github.com/en/copilot/how-tos/use-ai-models/change-the-chat-model#adding-more-models).
32
+
28
33
You can further extend the list of available models by [using your own language model API key](#bring-your-own-language-model-key).
29
34
30
35
If you have a paid Copilot plan, the model picker shows the premium request multiplier for premium models. Learn more about [premium requests](https://docs.github.com/en/copilot/managing-copilot/monitoring-usage-and-entitlements/about-premium-requests#premium-requests) in the GitHub Copilot documentation.
Copy file name to clipboardExpand all lines: docs/intelligentapps/agentbuilder.md
+54-65Lines changed: 54 additions & 65 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,6 @@
1
1
---
2
2
ContentId: bd3d7555-3d84-4500-ae95-6dcd39641af0
3
-
DateApproved: 07/14/2025
3
+
DateApproved: 10/03/2025
4
4
MetaDescription: Get Started with creating, iterating and optimizing your agents in AI Toolkit.
5
5
---
6
6
# Build agents and prompts in AI Toolkit
@@ -9,9 +9,8 @@ MetaDescription: Get Started with creating, iterating and optimizing your agents
9
9
> Agent Builder was previously known as Prompt Builder. The updated name better reflects the feature's capabilities and its focus on building agents.
10
10
11
11
Agent Builder in AI Toolkit streamlines the engineering workflow for building agents, including prompt engineering and integration with tools, such as MCP servers. It helps with common prompt engineering tasks:
12
-
- Generate starter prompts
13
-
- Iterate and refine with each run
14
-
- Break down complex tasks through prompt chaining and structured outputs
12
+
13
+
- Iterate and refine in real-time
15
14
- Provide easy access to code for seamless Large Language Model (LLM) integration via APIs
16
15
17
16
Agent Builder also enhances intelligent app's capabilities with tool use:
@@ -27,29 +26,27 @@ To access Agent Builder, use either of these options:
27
26
28
27
- In the AI Toolkit view, select **Agent Builder**
29
28
- Select **Try in Agent Builder** from a model card in the model catalog
29
+
- In the My Resources view, under **Models**, right-select a model and choose **Load in Agent Builder**
30
30
31
31
To test a prompt in Agent Builder, follow these steps:
32
32
33
-
1.In **Models**, select a modelfrom the dropdown list, or select **Browse models** to add another model from the model catalog.
33
+
1.If you haven't chosen a model, select one from the **Model**dropdown list in Agent Builder. You can also select **Browse models** to add a different model from the model catalog.
34
34
35
35
36
36
37
-
1. Enter a **User prompt** and optionally enter a **System prompt**.
38
-
39
-
The *user prompt* is the input that you want to send to the model. The optional *system prompt* is used to provide instructions with relevant context to guide the model response.
40
-
41
-
> [!TIP]
42
-
> Describe your project idea using natural language to generate prompts automatically.
43
-
> 
44
-
45
-
3. Select **Run** to send the prompts to the model.
37
+
1. Enter the agent instructions.
46
38
47
-
4. Optionally, select **Add Prompt** to add more prompts or **Add to Prompts**to build conversation history.
39
+
Use the **Instructions**field to tell your agent exactly what to do and how to do it. List the specific tasks, put them in order, and add any special instructions like tone or how to engage.
48
40
49
-
1. Repeat the previous steps to iterate over your prompts by observing the model response and making changes to the prompts.
41
+
1. Iterate over your instructions by observing the model response and making changes to the instructions.
42
+
1. Use the `{{your_variable}}` syntax to add a dynamic value in instructions. For example, add a variable called `user_name` and use it in your instructions like this: `Greet the user by their name: {{user_name}}`.
43
+
1. Provide a value for the variable in the **Variables** section.
44
+
1. Enter a prompt in the text box and select the send icon to test your agent.
45
+
1. Observe the model's response and make any necessary adjustments to your instructions.
50
46
51
47
## Use MCP servers
52
-
MCP server is a tool that allows you to connect to external APIs and services, enabling your agent to perform actions beyond just generating text. For example, you can use an MCP server to access databases, call web services, or interact with other applications.
48
+
49
+
An MCP server is a tool that allows you to connect to external APIs and services, enabling your agent to perform actions beyond just generating text. For example, you can use an MCP server to access databases, call web services, or interact with other applications.
53
50
54
51
Use the agent builder to discover and configure featured MCP servers, connect to existing MCP servers, or build a new MCP server from scaffold.
55
52
@@ -61,42 +58,56 @@ Use the agent builder to discover and configure featured MCP servers, connect to
61
58
AI Toolkit provides a list of featured MCP servers that you can use to connect to external APIs and services.
62
59
63
60
To configure an MCP server from featured selections, follow these steps:
64
-
1. In the **Tools** section, select **+ MCP Server**, and then select **+ Add server** in the Quick Pick.
65
-
2. Select **Use Featured MCP Servers** from the dropdown list.
66
-
3. Choose an MCP server that meets your needs.
67
-
68
-
4. Enter a name for the server.
69
-
5. Select tools you want to use.
61
+
1. In the **Tool** section, select **+ MCP Server**, and then select **MCP Server** in the Quick Pick.
62
+
1. Select **Could not find one? Browse more MCP servers** from the dropdown list.
63
+
64
+
65
+
1. Choose an MCP server that meets your needs.
66
+
67
+
68
+
1. The MCP server is added to your agent in the **MCP** subsection under **Tools**.
69
+
70
+
### Select tools from VS Code
71
+
72
+
1. In the **Tool** section, select **+ MCP Server**, and then select **MCP Server** in the Quick Pick.
73
+
1. Select **Use Tools Added in Visual Studio Code** from the dropdown list.
74
+
75
+
1. Select tools you want to use.
70
76
71
77
78
+
1. An MCP Server tool called `VSCode Tools` is added to your agent in the **MCP** subsection under **Tools**.
79
+
72
80
### Use an existing MCP server
73
81
> [!TIP]
74
82
> Find MCP servers in these [reference servers](https://github.com/modelcontextprotocol/servers?tab=readme-ov-file#-reference-servers).
75
83
76
84
To use an existing MCP server, follow these steps:
77
-
1. In the **Tools** section, select **+ MCP Server**, and then select **+ Add server** in the quick pick.
78
-
2. Select **Connect to an Existing MCP Server**
79
-
3. Select an option from the dropdown list to specify how you want to connect to the MCP server:
85
+
1. In the **MCP Workflow** section, select **+ Add MCP Server**.
86
+
87
+
1. Or in Agent Builder, in the **Tool** section, select the `+` icon to add a tool for your agent, and then select **+ Add server** in the Quick Pick.
88
+
1. Select **MCP server** in the Quick Pick.
89
+
1. Select **Connect to an Existing MCP Server**
90
+
1. Scroll down to the bottom of the dropdown list for the options to connect to the MCP server:
80
91
-**Command (stdio)**: Run a local command that implements the MCP protocol
81
-
-**HTTP (server-sent events)**: Connect to a remote server that implements the MCP protocol
82
-
4. Select tools from the MCP server if there are multiple tools available.
83
-
5. Enter your prompts and select **Run** to test the connection.
92
+
-**HTTP (HTTP or server-sent events)**: Connect to a remote server that implements the MCP protocol
93
+
1. Select tools from the MCP server if there are multiple tools available.
94
+
1. Enter your prompts in the text box and select the send icon to test the connection.
84
95
85
96
Here's an example of configuring the [Filesystem](https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem) server in AI Toolkit:
86
97
87
-
1. In the **Tools** section, select **+ MCP Server**, and then select **+ Add server** in the Quick Pick.
88
-
1. Select **Connect to an Existing MCP Server**
89
-
1.Select**Command (stdio)**
98
+
1. In the **Tool** section, select **+ MCP Server** in the Quick Pick.
99
+
1. Select **Could not find one? Browse more MCP servers** from the dropdown list.
100
+
1.Scroll down to the bottom of the dropdown list and select**Command (stdio)**
90
101
> [!NOTE]
91
102
> Some servers use the Python runtime and the `uvx` command. The process is the same as using the `npx` command.
92
103
1. Navigate to the [Server instructions](https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem#npx) and locate the `npx` section.
93
104
1. Copy the `command` and `args` into the input box in AI Toolkit. For the Filesystem server example, it's `npx -y @modelcontextprotocol/server-filesystem /Users/<username>/.aitk/examples`
94
-
1. Input a name for the server.
105
+
1. Input an ID for the server.
95
106
1. Optionally, enter extra environment variables.
96
107
Some servers might require extra environment variables such as API keys. In this case, AI Toolkit fails at the stage of adding tools and a file `mcp.json` opens, where you can enter the required server details following the instructions provided by each server.
97
108
98
109
After you complete the configuration:
99
-
1. Navigate back to **Tools** section and select **+ MCP Server**
110
+
1. Navigate back to **Tool** section and select **+ MCP Server**
100
111
1. Select the server you configured from the dropdown list
101
112
102
113
1. Select the tools you want to use.
@@ -106,39 +117,37 @@ Here's an example of configuring the [Filesystem](https://github.com/modelcontex
106
117
AI Toolkit also provides a scaffold to help you build a new MCP server. The scaffold includes a basic implementation of the MCP protocol, which you can customize to suit your needs.
107
118
108
119
### Build a new MCP server
120
+
109
121
To build a new MCP server, follow these steps:
110
-
1. In the **Tools** section, select **+ MCP Server**, and then select **+ Add server** in the quick pick.
111
-
1. Select **Create a New MCP Server**
122
+
1. In the **MCP Workflow** section, select **Create New MCP Server**.
112
123
1. Select a programming language from the dropdown list: **Python** or **TypeScript**
113
124
1. Select a folder to create the new MCP server project in.
114
125
1. Enter a name for the MCP server project.
115
126
116
-
117
-
118
127
After you create the MCP server project, you can customize the implementation to suit your needs. The scaffold includes a basic implementation of the MCP protocol, which you can modify to add your own functionality.
119
128
120
-
You can also use the agent builder to test the MCP server. The agent builder sends the prompts to the MCP server and displays the response.
129
+
You can also use the Agent Builder to test the MCP server. The Agent Builder sends the prompts to the MCP server and displays the response.
121
130
122
131
Follow these steps to test the MCP server:
123
132
124
133
> [!NOTE]
125
134
> To run the MCP Server in your local dev machine, you need: [Node.js](https://nodejs.org/) or Python installed on your machine.
126
135
127
136
1. Open VS Code Debug panel. Select `Debug in Agent Builder` or press `F5` to start debugging the MCP server.
128
-
1. Use AI Toolkit Agent Builder to test the server with the following prompt:
129
-
1. System Prompt: You are a weather forecast professional that can tell weather information based on given location.
130
137
1. The server is automatically connected to Agent Builder.
131
-
1. Select `Run` to test the server with the prompt.
138
+
1. Use AI Toolkit Agent Builder to enable the agent with the following instructions:
139
+
- "You are a weather forecast professional that can tell weather information based on given location.".
132
140
133
-
141
+
1. Enter the prompt "What is the weather in Seattle?" in the prompt box and select the send icon to test the server with the prompt.
142
+
1. Observe the response from the MCP server in the Agent Builder.
134
143
135
144
## Use function calling
136
145
137
146
Function calling connects your agent to external APIs and services.
138
147
139
148
140
149
141
-
1. In **Tools**, select **Add Tool**, then **Custom Tool**.
150
+
1. In **Tool**, select **Add Tool**, then **Custom Tool**.
142
151
1. Choose how to add the tool:
143
152
-**By Example**: Add from a JSON schema example
144
153
-**Upload Existing Schema**: Upload a JSON schema file
@@ -153,26 +162,6 @@ Use function calling tools in the **Evaluation** tab by entering mock responses
153
162
154
163
155
164
156
-
## Structured output
157
-
158
-
Structured output support helps you design prompts to deliver outputs in a structured, predictable format.
159
-
160
-
161
-
162
-
To test using structured output in Agent Builder, follow these steps:
163
-
164
-
1. Select the **Structure output** from the left area, and select **json_schema**.
165
-
166
-
1. Select **Prepare schema**, and then select **Select local file** to use your own schema, or select **Use an example** to use a predefined schema.
167
-
168
-
If you proceed with an example, you can select a schema from the dropdown list.
169
-
170
-
1. Select **Run** to send the prompts to the selected model.
171
-
172
-
1. You can also edit the schema by selecting name of the schema.
173
-
174
-
175
-
176
165
## Integrate prompt engineering into your application
177
166
178
167
After experimenting with models and prompts, you can get into coding right away with the automatically generated Python code.
@@ -185,7 +174,7 @@ To view the Python code, follow these steps:
185
174
186
175
1. For models hosted on GitHub, select the inference SDK you want to use.
187
176
188
-
AI Toolkit generates the code for the model you selected by using the provider's client SDK. For models hosted by GitHub, you can choose which inference SDK you want to use: [Azure AI Inference SDK](https://learn.microsoft.com/python/api/overview/azure/ai-inference-readme?view=azure-python-preview) or the SDK from the model provider, such as [OpenAI SDK](https://platform.openai.com/docs/libraries) or [Mistral API](https://docs.mistral.ai/api).
177
+
AI Toolkit generates the code for the model you selected by using the provider's client SDK. For models hosted by GitHub, you can choose which inference SDK you want to use: [Agent Framework SDK](https://github.com/microsoft/agent-framework) or the SDK from the model provider, such as [OpenAI SDK](https://platform.openai.com/docs/libraries) or [Mistral API](https://docs.mistral.ai/api).
189
178
190
179
1. The generated code snippet is shown in a new editor, where you can copy it into your application.
0 commit comments