bgauryy/octocode

What it solves

Octocode is an agentic research platform designed to move AI agents and developers from "guess-driven" to "research-driven" coding. It provides a unified toolset to gather concrete evidence from local workspaces and external sources (GitHub, npm) so that agents can base their changes and explanations on actual code structure and history rather than assumptions.

How it works

Octocode uses a high-performance Rust engine to provide a set of 12 tools accessible via a Command Line Interface (CLI) or as a Model Context Protocol (MCP) server. It integrates with local filesystems, GitHub APIs, and Language Server Protocol (LSP) servers to perform semantic navigation and structural analysis. To optimize token usage, it can minify or skeletonize code on the fly across 70+ languages, ensuring agents receive relevant structure without boilerplate noise.

Who it’s for

It is built for AI agents and developers who need to research large-scale codebases, compare patterns across multiple repositories, and perform deep semantic navigation (like resolving definitions and call hierarchies) across both local and remote code.

Highlights

• Unified Research Flow: Combines local search, GitHub repository/PR research, and npm package lookups in one toolset.
• MCP & CLI Support: Can be integrated directly into AI assistants (like Cursor, VS Code, Windsurf) or used as a standalone terminal tool.
• Semantic Navigation: Leverages LSP for typed navigation, including callers, callees, and type definitions.
• Structural Search: Supports AST-shaped searches using patterns or YAML rules to avoid false positives from comments or strings.
• Token Efficiency: Minifies and skeletonizes code to reduce the number of tokens spent on boilerplate.
• Built-in Security: Automatically scans and redacts secrets (AWS, OpenAI, etc.) from all content before it reaches the LLM.