Skip to content
All Skills

GitHub Copilot SDK Go Instructions

This file provides guidance on building Go applications using GitHub Copilot SDK.

Agent Orchestration|v1|Updated 7/14/2026|GitHub source
MCP get_skill({ skillId: "instructions-copilot-sdk-go-301b008a" })

Use this skill with your agent

Create a free account and connect via MCP

Get Started Free
## Core Principles

- The SDK is in technical preview and may have breaking changes
- Requires Go 1.21 or later
- Requires GitHub Copilot CLI installed and in PATH
- Uses goroutines and channels for concurrent operations
- No external dependencies beyond the standard library

## Installation

Always install via Go modules:

```bash
go get github.com/github/copilot-sdk/go
```

## Client Initialization

### Basic Client Setup

```go
import "github.com/github/copilot-sdk/go"

client := copilot.NewClient(nil)
if err := client.Start(); err != nil {
    log.Fatal(err)
}
defer client.Stop()
```

### Client Configuration Options

When creating a CopilotClient, use `ClientOptions`:

- `CLIPath` - Path to CLI executable (default: "copilot" from PATH)
- `CLIUrl` - URL of existing CLI server (e.g., "localhost:8080"). When provided, client won't spawn a process
- `Port` - Server port (default: 0 for random)
- `UseStdio` - Use stdio transport instead of TCP (default: true)
- `LogLevel` - Log level (default: "info")
- `AutoStart` - Auto-start server (default: true, use pointer: `boolPtr(true)`)
- `AutoRestart` - Auto-restart on crash (default: true, use pointer: `boolPtr(true)`)
- `Cwd` - Working directory for the CLI process
- `Env` - Environment variables for the CLI process ([]string)

### Manual Server Control

For explicit control:

```go
autoStart := false
client := copilot.NewClient(&copilot.ClientOptions{AutoStart: &autoStart})
if err := client.Start(); err != nil {
    log.Fatal(err)
}
// Use client...
client.Stop()
```

Use `ForceStop()` when `Stop()` takes too long.

## Session Management

### Creating Sessions

Use `SessionConfig` for configuration:

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Model: "gpt-5",
    Streaming: true,
    Tools: []copilot.Tool{...},
    SystemMessage: &copilot.SystemMessageConfig{ ... },
    AvailableTools: []string{"tool1", "tool2"},
    ExcludedTools: []string{"tool3"},
    Provider: &copilot.ProviderConfig{ ... },
})
if err != nil {
    log.Fatal(err)
}
```

### Session Config Options

- `SessionID` - Custom session ID
- `Model` - Model name ("gpt-5", "claude-sonnet-4.5", etc.)
- `Tools` - Custom tools exposed to the CLI ([]Tool)
- `SystemMessage` - System message customization (\*SystemMessageConfig)
- `AvailableTools` - Allowlist of tool names ([]string)
- `ExcludedTools` - Blocklist of tool names ([]string)
- `Provider` - Custom API provider configuration (BYOK) (\*ProviderConfig)
- `Streaming` - Enable streaming response chunks (bool)
- `MCPServers` - MCP server configurations
- `CustomAgents` - Custom agent configurations
- `ConfigDir` - Config directory override
- `SkillDirectories` - Skill directories ([]string)
- `DisabledSkills` - Disabled skills ([]string)

### Resuming Sessions

```go
session, err := client.ResumeSession("session-id", &copilot.ResumeSessionConfig{OnPermissionRequest: copilot.PermissionHandler.ApproveAll})
// Or with options:
session, err := client.ResumeSessionWithOptions("session-id", &copilot.ResumeSessionConfig{ ... })
```

### Session Operations

- `session.SessionID` - Get session identifier (string)
- `session.Send(copilot.MessageOptions{Prompt: "...", Attachments: []copilot.Attachment{...}})` - Send message, returns (messageID string, error)
- `session.SendAndWait(options, timeout)` - Send and wait for idle, returns (\*SessionEvent, error)
- `session.Abort()` - Abort current processing, returns error
- `session.GetMessages()` - Get all events/messages, returns ([]SessionEvent, error)
- `session.Destroy()` - Clean up session, returns error

## Event Handling

### Event Subscription Pattern

ALWAYS use channels or done signals to wait for session events:

```go
done := make(chan struct{})

unsubscribe := session.On(func(evt copilot.SessionEvent) {
    switch evt.Type {
    case copilot.AssistantMessage:
        fmt.Println(*evt.Data.Content)
    case copilot.SessionIdle:
        close(done)
    }
})
defer unsubscribe()

session.Send(copilot.MessageOptions{Prompt: "..."})
<-done
```

### Unsubscribing from Events

The `On()` method returns a function that unsubscribes:

```go
unsubscribe := session.On(func(evt copilot.SessionEvent) {
    // handler
})
// Later...
unsubscribe()
```

### Event Types

Use type switches for event handling:

```go
session.On(func(evt copilot.SessionEvent) {
    switch evt.Type {
    case copilot.UserMessage:
        // Handle user message
    case copilot.AssistantMessage:
        if evt.Data.Content != nil {
            fmt.Println(*evt.Data.Content)
        }
    case copilot.ToolExecutionStart:
        // Tool execution started
    case copilot.ToolExecutionComplete:
        // Tool execution completed
    case copilot.SessionStart:
        // Session started
    case copilot.SessionIdle:
        // Session is idle (processing complete)
    case copilot.SessionError:
        if evt.Data.Message != nil {
            fmt.Println("Error:", *evt.Data.Message)
        }
    }
})
```

## Streaming Responses

### Enabling Streaming

Set `Streaming: true` in SessionConfig:

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Model: "gpt-5",
    Streaming: true,
})
```

### Handling Streaming Events

Handle both delta events (incremental) and final events:

```go
done := make(chan struct{})

session.On(func(evt copilot.SessionEvent) {
    switch evt.Type {
    case copilot.AssistantMessageDelta:
        // Incremental text chunk
        if evt.Data.DeltaContent != nil {
            fmt.Print(*evt.Data.DeltaContent)
        }
    case copilot.AssistantReasoningDelta:
        // Incremental reasoning chunk (model-dependent)
        if evt.Data.DeltaContent != nil {
            fmt.Print(*evt.Data.DeltaContent)
        }
    case copilot.AssistantMessage:
        // Final complete message
        fmt.Println("\n--- Final ---")
        if evt.Data.Content != nil {
            fmt.Println(*evt.Data.Content)
        }
    case copilot.AssistantReasoning:
        // Final reasoning content
        fmt.Println("--- Reasoning ---")
        if evt.Data.Content != nil {
            fmt.Println(*evt.Data.Content)
        }
    case copilot.SessionIdle:
        close(done)
    }
})

session.Send(copilot.MessageOptions{Prompt: "Tell me a story"})
<-done
```

Note: Final events (`AssistantMessage`, `AssistantReasoning`) are ALWAYS sent regardless of streaming setting.

## Custom Tools

### Defining Tools

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Model: "gpt-5",
    Tools: []copilot.Tool{
        {
            Name:        "lookup_issue",
            Description: "Fetch issue details from tracker",
            Parameters: map[string]interface{}{
                "type": "object",
                "properties": map[string]interface{}{
                    "id": map[string]interface{}{
                        "type":        "string",
                        "description": "Issue ID",
                    },
                },
                "required": []string{"id"},
            },
            Handler: func(inv copilot.ToolInvocation) (copilot.ToolResult, error) {
                args := inv.Arguments.(map[string]interface{})
                issueID := args["id"].(string)

                issue, err := fetchIssue(issueID)
                if err != nil {
                    return copilot.ToolResult{}, err
                }

                return copilot.ToolResult{
                    TextResultForLLM: fmt.Sprintf("Issue: %v", issue),
                    ResultType:       "success",
                    ToolTelemetry:    map[string]interface{}{},
                }, nil
            },
        },
    },
})
```

### Tool Return Types

- Return `ToolResult` struct with fields:
  - `TextResultForLLM` (string) - Result text for the LLM
  - `ResultType` (string) - "success" or "failure"
  - `Error` (string, optional) - Internal error message (not shown to LLM)
  - `ToolTelemetry` (map[string]interface{}) - Telemetry data

### Tool Execution Flow

When Copilot invokes a tool, the client automatically:

1. Runs your handler function
2. Returns the ToolResult
3. Responds to the CLI

## System Message Customization

### Append Mode (Default - Preserves Guardrails)

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Model: "gpt-5",
    SystemMessage: &copilot.SystemMessageConfig{
        Mode: "append",
        Content: `
<workflow_rules>
- Always check for security vulnerabilities
- Suggest performance improvements when applicable
</workflow_rules>
`,
    },
})
```

### Replace Mode (Full Control - Removes Guardrails)

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Model: "gpt-5",
    SystemMessage: &copilot.SystemMessageConfig{
        Mode:    "replace",
        Content: "You are a helpful assistant.",
    },
})
```

## File Attachments

Attach files to messages using `Attachment`:

```go
messageID, err := session.Send(copilot.MessageOptions{
    Prompt: "Analyze this file",
    Attachments: []copilot.Attachment{
        {
            Type:        "file",
            Path:        "/path/to/file.go",
            DisplayName: "My File",
        },
    },
})
```

## Message Delivery Modes

Use the `Mode` field in `MessageOptions`:

- `"enqueue"` - Queue message for processing
- `"immediate"` - Process message immediately

```go
session.Send(copilot.MessageOptions{
    Prompt: "...",
    Mode:   "enqueue",
})
```

## Multiple Sessions

Sessions are independent and can run concurrently:

```go
session1, _ := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
	Model:               "gpt-5",
})
session2, _ := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
	Model:               "claude-sonnet-4.5",
})

session1.Send(copilot.MessageOptions{Prompt: "Hello from session 1"})
session2.Send(copilot.MessageOptions{Prompt: "Hello from session 2"})
```

## Bring Your Own Key (BYOK)

Use custom API providers by configuring `ProviderConfig`:

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Provider: &copilot.ProviderConfig{
        Type:    "openai",
        BaseURL: "https://api.openai.com/v1",
        APIKey:  "your-api-key",
    },
})
```

## Session Lifecycle Management

### Checking Connection State

```go
state := client.GetState()
// Returns: "disconnected", "connecting", "connected", or "error"
```

## Error Handling

### Standard Exception Handling

```go
session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
})
if err != nil {
    log.Fatalf("Failed to create session: %v", err)
}

_, err = session.Send(copilot.MessageOptions{Prompt: "Hello"})
if err != nil {
    log.Printf("Failed to send: %v", err)
}
```

### Session Error Events

Monitor `SessionError` type for runtime errors:

```go
session.On(func(evt copilot.SessionEvent) {
    if evt.Type == copilot.SessionError {
        if evt.Data.Message != nil {
            fmt.Fprintf(os.Stderr, "Session Error: %s\n", *evt.Data.Message)
        }
    }
})
```

## Connectivity Testing

Use `Ping` to verify server connectivity:

```go
resp, err := client.Ping("test message")
if err != nil {
    log.Printf("Server unreachable: %v", err)
} else {
    log.Printf("Server responded at %d", resp.Timestamp)
}
```

## Resource Cleanup

### Cleanup with Defer

ALWAYS use `defer` for cleanup:

```go
client := copilot.NewClient(nil)
if err := client.Start(); err != nil {
    log.Fatal(err)
}
defer client.Stop()

session, err := client.CreateSession(&copilot.SessionConfig{OnPermissionRequest: copilot.PermissionHandler.ApproveAll})
if err != nil {
    log.Fatal(err)
}
defer session.Destroy()
```

### Manual Cleanup

If not using defer:

```go
client := copilot.NewClient(nil)
err := client.Start()
if err != nil {
    log.Fatal(err)
}

session, err := client.CreateSession(&copilot.SessionConfig{OnPermissionRequest: copilot.PermissionHandler.ApproveAll})
if err != nil {
    client.Stop()
    log.Fatal(err)
}

// Use session...

session.Destroy()
errors := client.Stop()
for _, err := range errors {
    log.Printf("Cleanup error: %v", err)
}
```

## Best Practices

1. **Always use `defer`** for cleanup of clients and sessions
2. **Use channels** to wait for SessionIdle event
3. **Handle SessionError** events for robust error handling
4. **Use type switches** for event handling
5. **Enable streaming** for better UX in interactive scenarios
6. **Provide descriptive tool names and descriptions** for better model understanding
7. **Call unsubscribe functions** when no longer needed
8. **Use SystemMessageConfig with Mode: "append"** to preserve safety guardrails
9. **Handle both delta and final events** when streaming is enabled
10. **Check nil pointers** in event data (Content, Message, etc. are pointers)

## Common Patterns

### Simple Query-Response

```go
client := copilot.NewClient(nil)
if err := client.Start(); err != nil {
    log.Fatal(err)
}
defer client.Stop()

session, err := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
	Model:               "gpt-5",
})
if err != nil {
    log.Fatal(err)
}
defer session.Destroy()

done := make(chan struct{})

session.On(func(evt copilot.SessionEvent) {
    if evt.Type == copilot.AssistantMessage && evt.Data.Content != nil {
        fmt.Println(*evt.Data.Content)
    } else if evt.Type == copilot.SessionIdle {
        close(done)
    }
})

session.Send(copilot.MessageOptions{Prompt: "What is 2+2?"})
<-done
```

### Multi-Turn Conversation

```go
session, _ := client.CreateSession(&copilot.SessionConfig{OnPermissionRequest: copilot.PermissionHandler.ApproveAll})
defer session.Destroy()

sendAndWait := func(prompt string) error {
    done := make(chan struct{})
    var eventErr error

    unsubscribe := session.On(func(evt copilot.SessionEvent) {
        switch evt.Type {
        case copilot.AssistantMessage:
            if evt.Data.Content != nil {
                fmt.Println(*evt.Data.Content)
            }
        case copilot.SessionIdle:
            close(done)
        case copilot.SessionError:
            if evt.Data.Message != nil {
                eventErr = fmt.Errorf(*evt.Data.Message)
            }
        }
    })
    defer unsubscribe()

    if _, err := session.Send(copilot.MessageOptions{Prompt: prompt}); err != nil {
        return err
    }
    <-done
    return eventErr
}

sendAndWait("What is the capital of France?")
sendAndWait("What is its population?")
```

### SendAndWait Helper

```go
// Use built-in SendAndWait for simpler synchronous interaction
response, err := session.SendAndWait(copilot.MessageOptions{
    Prompt: "What is 2+2?",
}, 0) // 0 uses default 60s timeout

if err != nil {
    log.Printf("Error: %v", err)
}
if response != nil && response.Data.Content != nil {
    fmt.Println(*response.Data.Content)
}
```

### Tool with Struct Return Type

```go
type UserInfo struct {
    ID    string `json:"id"`
    Name  string `json:"name"`
    Email string `json:"email"`
    Role  string `json:"role"`
}

session, _ := client.CreateSession(&copilot.SessionConfig{
	OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
    Tools: []copilot.Tool{
        {
            Name:        "get_user",
            Description: "Retrieve user information",
            Parameters: map[string]interface{}{
                "type": "object",
                "properties": map[string]interface{}{
                    "user_id": map[string]interface{}{
                        "type":        "string",
                        "description": "User ID",
                    },
                },
                "required": []string{"user_id"},
            },
            Handler: func(inv copilot.ToolInvocation) (copilot.ToolResult, error) {
                args := inv.Arguments.(map[string]interface{})
                userID := args["user_id"].(string)

                user := UserInfo{
                    ID:    userID,
                    Name:  "John Doe",
                    Email: "john@example.com",
                    Role:  "Developer",
                }

                jsonBytes, _ := json.Marshal(user)
                return copilot.ToolResult{
                    TextResultForLLM: string(jsonBytes),
                    ResultType:       "success",
                    ToolTelemetry:    map[string]interface{}{},
                }, nil
            },
        },
    },
})
```
#github-copilot#m365#copilot#agentscopilot-cligo

Related Skills

More skills in Agent Orchestration

1. Product type search — what design patterns fit this product?

Design UI/UX systems with style guides, palettes, typography, and component specs for new interfaces

#github#broad-capabilityMIT

Acp Router

Route plain-language requests for Claude Code, Cursor, Copilot, OpenClaw ACP, OpenCode, Gemini CLI, Qwen, Kiro, Kimi, iFlow, Factory Droid, Kilocode, or explicit ACP harness work into either OpenClaw ACP runtime sessions or direct acpx-driven sessions ("telephone game" flow). For coding-agent thread requests, read this skill first, then use only `sessions_spawn` for thread creation. Codex chat binding defaults to the native Codex app-server plugin unless ACP is explicit or background spawn needs ACP.

#broad-capability#browserMIT

Add New Safe Output Type

Adding a New Safe Output Type to GitHub Agentic Workflows

#github#broad-capabilityMIT

Agent Governance

Patterns and techniques for adding governance, safety, and trust controls to AI agent systems. Use this skill when: - Building AI agents that call external tools (APIs, databases, file systems) - Implementing policy-based access controls for agent tool usage - Adding semantic intent classification to detect dangerous prompts - Creating trust scoring systems for multi-agent workflows - Building audit trails for agent actions and decisions - Enforcing rate limits, content filters, or tool restrictions on agents - Working with any agent framework (PydanticAI, CrewAI, OpenAI Agents, LangChain, AutoGen)

#github-copilot#agentMIT

Agentic Development

Build AI agents with Pydantic AI (Python) and Claude SDK (Node.js)

#claude-bootstrap#bootstrapMIT

Agentic Workflows

Route gh-aw workflow design/create/debug/upgrade requests to the right prompts.

#broad-capability#agentic-workflowsMIT