Delve MCP Server

Overview

The Go Delve Debugger MCP Server bridges the power of the Model Context Protocol with the widely used Delve debugger for Go. By exposing a rich set of debugging tools over MCP, it lets large language models and other AI assistants control real Go processes as if they were local development tools. This removes the need for developers to manually start, attach, and step through code; instead an AI can reason about program state, suggest fixes, or even generate test cases on the fly.

At its core, the server runs a lightweight MCP service that translates JSON‑RPC calls into DAP (Debug Adapter Protocol) commands sent to a Delve instance. Each debug session is uniquely identified, allowing multiple concurrent runs—ideal for multi‑project environments or continuous integration pipelines where parallel debugging is required. The server manages the lifecycle of these sessions, from creation with to clean termination via , and offers introspection tools such as .

Key capabilities include:

• Breakpoint Management – set, remove, and query breakpoints by file and line.
• Execution Control – , , , and let the AI steer execution flow precisely.
• Expression Evaluation – inspect variables, evaluate arbitrary Go expressions, and access specific stack frames with .
• Session Multiplexing – start several debug sessions simultaneously, each isolated yet accessible through the same MCP endpoint.

These features make the server invaluable for scenarios such as:

• AI‑driven Pair Programming – an assistant can debug a user’s code, suggest breakpoint placements, and explain variable values in natural language.
• Automated Test Generation – a model can run tests, capture stack traces on failure, and propose corrective code snippets.
• Continuous Integration Debugging – during a CI run, an AI can pause failing builds, inspect state, and recommend fixes without manual intervention.

Integration is straightforward for any MCP‑aware workflow. A client (e.g., Cursor, a custom tool, or the provided demo) connects to , sends tool calls defined in the server’s schema, and receives structured responses. The MCP server then forwards these to Delve, returning the results back through the same channel. This seamless pipeline means developers can embed deep debugging logic into chat‑based assistants, automated scripts, or IDE extensions without writing any custom debugger glue code.

In summary, the Go Delve Debugger MCP Server turns a command‑line debugger into a first‑class AI service. It empowers developers to leverage conversational models for real‑time, context‑aware debugging while preserving the full fidelity of Delve’s capabilities.