⚠️ Disclaimer: This is an unofficial, community-driven SDK and is not supported or endorsed by GitHub. Use at your own risk.
This guide covers advanced scenarios for extending and customizing your Copilot integration.
Let the AI call back into your application to fetch data or perform actions.
// Define strongly-typed arguments with a record
record IssueArgs(String id) {}
var lookupTool = ToolDefinition.create(
"lookup_issue",
"Fetch issue details from our tracker",
Map.of(
"type", "object",
"properties", Map.of(
"id", Map.of("type", "string", "description", "Issue identifier")
),
"required", List.of("id")
),
invocation -> {
IssueArgs args = invocation.getArgumentsAs(IssueArgs.class);
return CompletableFuture.completedFuture(fetchIssue(args.id()));
}
);
var session = client.createSession(
new SessionConfig()
.setTools(List.of(lookupTool))
).get();See ToolDefinition Javadoc for schema details.
Customize the AI's behavior by adding rules or replacing the default prompt.
Use APPEND mode to add constraints while keeping default guardrails:
var session = client.createSession(
new SessionConfig()
.setSystemMessage(new SystemMessageConfig()
.setMode(SystemMessageMode.APPEND)
.setContent("""
<rules>
- Always check for security vulnerabilities
- Suggest performance improvements
</rules>
"""))
).get();Use REPLACE mode for complete control (removes default guardrails):
var session = client.createSession(
new SessionConfig()
.setSystemMessage(new SystemMessageConfig()
.setMode(SystemMessageMode.REPLACE)
.setContent("You are a helpful coding assistant."))
).get();Include files as context for the AI to analyze.
session.send(new MessageOptions()
.setPrompt("Review this file for bugs")
.setAttachments(List.of(
new Attachment()
.setType("file")
.setPath("/path/to/file.java")
.setDisplayName("MyService.java")
))
).get();Use your own OpenAI or Azure OpenAI API key instead of GitHub Copilot.
var session = client.createSession(
new SessionConfig()
.setProvider(new ProviderConfig()
.setType("openai")
.setBaseUrl("https://api.openai.com/v1")
.setApiKey("sk-..."))
).get();Run long conversations without hitting context limits.
When enabled (default), the session automatically compacts older messages as the context window fills up.
var session = client.createSession(
new SessionConfig()
.setInfiniteSessions(new InfiniteSessionConfig()
.setEnabled(true)
.setBackgroundCompactionThreshold(0.80) // Start compacting at 80%
.setBufferExhaustionThreshold(0.95)) // Block at 95%
).get();
// Access the workspace where session state is persisted
String workspace = session.getWorkspacePath();When compaction occurs, the session emits events that you can listen for:
session.on(event -> {
if (event instanceof SessionCompactionStartEvent start) {
System.out.println("Compaction started");
} else if (event instanceof SessionCompactionCompleteEvent complete) {
var data = complete.getData();
System.out.println("Compaction completed - success: " + data.isSuccess()
+ ", tokens removed: " + data.getTokensRemoved());
}
});For short conversations, disable to avoid overhead:
new InfiniteSessionConfig().setEnabled(false)Extend the AI with external tools via the Model Context Protocol.
Map<String, Object> server = Map.of(
"type", "local",
"command", "npx",
"args", List.of("-y", "@modelcontextprotocol/server-filesystem", "/tmp"),
"tools", List.of("*")
);
var session = client.createSession(
new SessionConfig()
.setMcpServers(Map.of("filesystem", server))
).get();📖 Full MCP documentation → for local/remote servers and all options.
Load custom skills from directories to extend the AI's capabilities with domain-specific knowledge.
Skills are loaded from SKILL.md files in subdirectories of the specified skill directories:
var session = client.createSession(
new SessionConfig()
.setSkillDirectories(List.of("/path/to/skills"))
).get();Each skill subdirectory should contain a SKILL.md file with YAML frontmatter:
---
name: my-skill
description: A skill that provides domain-specific knowledge
---
# Skill Instructions
Your skill instructions go here...Disable specific skills by name:
var session = client.createSession(
new SessionConfig()
.setSkillDirectories(List.of("/path/to/skills"))
.setDisabledSkills(List.of("my-skill"))
).get();Use a custom configuration directory for session settings:
var session = client.createSession(
new SessionConfig()
.setConfigDir("/path/to/custom/config")
).get();This is useful when you need to isolate session configuration or use different settings for different environments.
Approve or deny permission requests from the AI.
var session = client.createSession(
new SessionConfig()
.setOnPermissionRequest((request, invocation) -> {
// Inspect request and approve/deny
var result = new PermissionRequestResult();
result.setKind("user-approved");
return CompletableFuture.completedFuture(result);
})
).get();Control the CLI lifecycle yourself instead of auto-start.
var client = new CopilotClient(
new CopilotClientOptions().setAutoStart(false)
);
client.start().get(); // Start manually
// ... use client ...
client.stop().get(); // Stop manuallyAll SDK methods return CompletableFuture. Errors surface via ExecutionException:
try {
session.send(new MessageOptions().setPrompt("Hello")).get();
} catch (ExecutionException ex) {
System.err.println("Error: " + ex.getCause().getMessage());
}For reactive error handling, use exceptionally() or handle():
session.send(new MessageOptions().setPrompt("Hello"))
.exceptionally(ex -> {
System.err.println("Failed: " + ex.getMessage());
return null;
});