Make CoPilot an Interactive Programmer

What is Make CoPilot an Interactive Programmer?

Make CoPilot an Interactive Programmer is a VS Code extension (Calva Backseat Driver) that gives AI agents—such as GitHub Copilot and Cursor—access to Calva’s Clojure tools, including the REPL, file management, and structural editing. It is designed for Clojure developers who want to use AI for interactive programming directly within VS Code.

How to use Make CoPilot an Interactive Programmer?

Install the Calva and Calva Backseat Driver extensions in VS Code, then connect Calva to your Clojure/ClojureScript project. For Copilot and Cursor, zero configuration is required—simply ask Copilot to use the REPL. For other AI harnesses, configure the MCP server per project. The agent will automatically discover and use the provided tools and skills.

Key features of Make CoPilot an Interactive Programmer

• Evaluate Clojure code via the REPL in any active session
• Create Clojure files with automatic bracket balancing
• Append code to existing Clojure files with bracket balancing
• Replace top-level forms with structural editing and linting
• Insert top-level forms with formatting and bracket balancing
• Look up symbol documentation, arguments, and ClojureDocs examples

Use cases of Make CoPilot an Interactive Programmer

• AI-assisted REPL-driven development in Clojure projects
• Structural editing of Clojure code via natural language commands
• Parallel subagents investigating implementation approaches simultaneously
• Learning Clojure idioms by having the AI look up symbols and docs
• Rapid prototyping with zero-config Copilot integration

FAQ from Make CoPilot an Interactive Programmer

What dependencies are required?

VS Code, the Calva extension, the Calva Backseat Driver extension, and GitHub Copilot (or another MCP-compliant assistant).

How does the server connect to AI agents?

For Copilot and Cursor, it uses VS Code LM Tools and the Cursor MCP API with zero configuration. Other agents require manual MCP setup.

Are there limits on REPL evaluation results?

Yes: calva-backseat-driver.evaluation.maxLength (default 25 items) and maxDepth (default 7 levels) prevent large results from overflowing the agent’s context window.

Where are project or user data stored?

The README does not specify data storage; all evaluation and editing happens live in the connected REPL and VS Code workspace.

Can I review edits made by the AI?

There is no built-in UI for reviewing edits; the README suggests using VS Code’s source control and Timeline view for this purpose.