Does the Copilot plugin for JetBrains provide an API for OTHER plugins to initiate a chat with an arbitrary prompt

[jfree]

Select Topic Area

Question

Copilot Feature Area

JetBrains & Xcode

Body

Hello, I'm working on a plugin for JetBrains IDEs and I want to know if it is possible for my plugin to programmatically launch a chat in the Copilot plugin? This would open up some very interesting integrations. Any pointers much appreciated!

[AmirhosseinHonardoust]

Hi

right now, the copilot plugin for JetBrains doesn’t provide a public API to programmatically open a chat or prefill a prompt. the supported way to integrate is through MCP or copilot extensions (if enabled), which let your plugin expose tools and context inside copilot chat.
you could try calling the IDE action to show the copilot chat window, but that’s undocumented and may break anytime. for anything stable, MCP is the recommended path.

[Aldayanday1]

🤖 Hey @jfree!

Great question! I totally understand wanting to integrate with Copilot Chat programmatically — that would unlock some amazing possibilities for plugin developers. Let me break down what's currently available and the best path forward! 💡

---

📋 TL;DR - Current State

Unfortunately, GitHub Copilot for JetBrains doesn't expose a public API for other plugins to programmatically open chat or prefill prompts yet. However, there are a couple of approaches you can take:

---

🛠️ Available Options

✅ Option 1: Model Context Protocol (MCP) - Recommended

The official and stable way to integrate with Copilot is through the Model Context Protocol (MCP). This lets your plugin:

• Expose custom tools to Copilot
• Provide additional context to Copilot Chat
• Let users invoke your plugin's functionality through natural language

📚 How MCP Works:

Your Plugin (MCP Server)
    ↓
  Provides tools/context
    ↓
Copilot Chat (MCP Client)
    ↓
  Uses your tools in responses

🚀 Quick Setup Guide:

Step 1: Implement MCP Server in Your Plugin

// Example: Expose a tool to Copilot
class MyMCPServer : MCPServer {
    override fun getTools(): List<Tool> {
        return listOf(
            Tool(
                name = "analyze_code",
                description = "Analyzes code complexity",
                inputSchema = JsonSchema(/* define params */)
            )
        )
    }
    
    override fun executeTool(name: String, args: JsonObject): ToolResult {
        when (name) {
            "analyze_code" -> {
                // Your plugin logic here
                return ToolResult.success("Analysis complete!")
            }
        }
    }
}

Step 2: Register Your MCP Server

<!-- plugin.xml -->
<extensions defaultExtensionNs="com.github.copilot">
    <mcpServer implementation="com.yourplugin.MyMCPServer"/>
</extensions>

Step 3: Users Can Now Invoke Your Plugin Through Copilot

User in Copilot Chat: "@workspace analyze the code complexity of this file"
                          ↓
                   Copilot uses your tool
                          ↓
                  Your plugin executes
                          ↓
                  Result shown in chat

📖 Learn More:

• Model Context Protocol Docs
• GitHub Copilot Extensions Guide

---

⚠️ Option 2: IDE Action Invocation - Unstable/Hacky

You could try invoking the internal IDE action to show Copilot Chat, but this is undocumented and may break in future updates:

// NOT RECOMMENDED - May break anytime
fun openCopilotChat() {
    val actionManager = ActionManager.getInstance()
    val action = actionManager.getAction("github.copilot.chat.show")
    
    if (action != null) {
        val event = AnActionEvent.createFromDataContext(
            ActionPlaces.UNKNOWN,
            null,
            dataContext
        )
        action.actionPerformed(event)
    }
}

Problems with this approach:

• ❌ No public API contract
• ❌ Action IDs may change
• ❌ Can't prefill prompts
• ❌ May break with plugin updates

---

🎯 Step-by-Step: Integrate with MCP

1️⃣ Add Copilot SDK Dependency

// build.gradle.kts
dependencies {
    implementation("com.github.copilot:copilot-sdk:1.0.0") // Check latest version
}

2️⃣ Create Your MCP Server Class

package com.yourplugin.copilot

import com.github.copilot.mcp.*

class CodeAnalyzerMCPServer : MCPServer {
    
    override val name = "code-analyzer"
    override val version = "1.0.0"
    
    override fun getTools(): List<Tool> {
        return listOf(
            Tool(
                name = "analyze_complexity",
                description = "Analyzes code complexity metrics",
                inputSchema = mapOf(
                    "type" to "object",
                    "properties" to mapOf(
                        "filePath" to mapOf("type" to "string")
                    )
                )
            ),
            Tool(
                name = "suggest_refactoring",
                description = "Suggests refactoring improvements",
                inputSchema = mapOf(
                    "type" to "object",
                    "properties" to mapOf(
                        "code" to mapOf("type" to "string")
                    )
                )
            )
        )
    }
    
    override suspend fun executeTool(
        name: String,
        arguments: Map<String, Any>
    ): ToolResult {
        return when (name) {
            "analyze_complexity" -> {
                val filePath = arguments["filePath"] as? String
                val analysis = performComplexityAnalysis(filePath)
                ToolResult.success(analysis)
            }
            "suggest_refactoring" -> {
                val code = arguments["code"] as? String
                val suggestions = generateRefactoringSuggestions(code)
                ToolResult.success(suggestions)
            }
            else -> ToolResult.error("Unknown tool: $name")
        }
    }
    
    private fun performComplexityAnalysis(filePath: String?): String {
        // Your analysis logic
        return "Complexity: Medium (score: 12)"
    }
    
    private fun generateRefactoringSuggestions(code: String?): String {
        // Your refactoring logic
        return "Suggestion: Extract method for lines 10-25"
    }
}

3️⃣ Register in plugin.xml

<idea-plugin>
    <extensions defaultExtensionNs="com.github.copilot">
        <mcpServer 
            implementation="com.yourplugin.copilot.CodeAnalyzerMCPServer"/>
    </extensions>
</idea-plugin>

4️⃣ Test Integration

1. Build and run your plugin
2. Open Copilot Chat in JetBrains
3. Type: @workspace analyze complexity of Main.java
4. Copilot will discover and use your tool!

---

💡 What You Can Build with MCP

Here are some cool integration ideas:

• 🔍 Code Analysis Tools - Expose linters, formatters, complexity analyzers
• 📊 Database Query Helpers - Let Copilot query your DB plugin
• 🐛 Debug Assistants - Provide debugging context to Copilot
• 📝 Documentation Generators - Generate docs through natural language
• 🧪 Test Generators - Create tests based on selected code

---

🔮 Future Possibilities

While there's no public API yet to programmatically open chat with arbitrary prompts, GitHub is actively working on:

• 📢 More extension points for Copilot
• 🔧 Better plugin integration APIs
• 🌐 Copilot Extensions ecosystem

Keep an eye on:

• GitHub Copilot Extensions Docs
• JetBrains Plugin SDK Updates

---

📚 Additional Resources

Resource	Link
MCP Specification	modelcontextprotocol.io
Copilot Extensions	GitHub Docs
JetBrains Plugin DevKit	JetBrains Docs
Copilot API Discussion	GitHub Community

---

🎬 Bottom Line

For a stable, future-proof integration:
✅ Use Model Context Protocol (MCP)
✅ Expose your plugin's capabilities as tools
✅ Let Copilot discover and invoke your functionality

For a quick hack (not recommended):
⚠️ Invoke internal IDE actions (may break)

---

Hope this helps you build something awesome!

Have a nice day jfree! ✨

[jfree]

Thanks for all the detail, much appreciated!