Alucify

Analyze your specs and codebase to surface risks before your AI coding agents start writing code.

Alucify builds structured project context from your PRDs, design docs, and source code. It finds gaps, conflicts, and missing requirements so Claude Code produces accurate code on the first pass instead of hallucinating.

Why Alucify?

AI coding agents work best when they have clear, complete context. But specs drift from code, requirements contradict each other, and coverage gaps hide in plain sight. Alucify catches these problems before you start coding.

• Fewer hallucinations — Agents get structured context instead of ambiguous docs
• Catch gaps early — Find missing requirements, untested flows, and spec conflicts before they become bugs
• Track progress — See which requirements are implemented, in progress, or still planned

Features

Build Project Context

Analyze your codebase and specification documents to build structured project context covering your requirements, implementations, and their relationships.

View AI Coding Risks

See a prioritized list of risks that would cause AI agents to write incorrect code — gaps between specs and implementation, conflicting requirements, missing validation criteria.

Refine Specs

Select risks and fix them interactively via an AI chat session. Alucify proposes spec updates, you approve or edit, and the fixes are applied directly to your documents.

Generate Specs

Create PRDs, architecture docs, research reports, UX designs, and more via interactive AI workflows (requires Claude Code CLI):

Analyze Progress

Track implementation progress against your requirements by analyzing git commits. See which features are implemented, in progress, or still planned.

Technical Debt Analysis

Run stress tests against your codebase architecture to identify structural weaknesses, coupling hotspots, and maintainability risks.

Token Usage Tracking

Monitor your AI token consumption across the project. Tracks both Alucify workflow usage and direct Claude Code sessions, with breakdowns by source, model, user, and time period. Available in the VS Code extension, CLI (alucify usage), and updated automatically on git commits.

Ask About Project

Chat with AI about your project using project context. Ask questions about architecture, requirements, or implementation details.

Language & Framework Support

Alucify uses Tree-sitter AST parsing for accurate code indexing:

Full = Tree-sitter AST + multi-tier call resolution + framework-specific classification. Supported = Tree-sitter AST + call resolution. Generic classification heuristics. Partial = File path heuristics only. No AST parsing. Planned = Tree-sitter grammar available. Not yet implemented.

Getting Started

Prerequisites

• VS Code 1.85.0+
• Node.js 18+
• One of:
  • Anthropic API key — pay-per-use via the Anthropic API
  • Claude Code CLI — use your existing Claude subscription

Quick Start

1. Install Alucify from the VS Code Marketplace
2. Open the Alucify sidebar (click the icon in the activity bar)
3. Complete the welcome setup — choose your auth mode and sign in
4. Click Analyze Code to index your codebase (free, no LLM calls)
5. Add spec documents via the Add Spec button (PDFs, Markdown, text files)
6. Click Analyze Specs to build project context and surface risks
7. Review findings in the AI Coding Risks panel

The .alucify/ Directory

All analysis output is stored in .alucify/ at your workspace root:

.alucify/
  appgraph-baseline.json    # Codebase context (from Analyze Code)
  appgraph-project.json     # Full project context (code + specs)
  artifact-manifest.json    # Spec document index
  artifacts/                # Your spec documents
  versions/                 # Version snapshots (baseline, specs, progress)
  progress/                 # Progress tracking checkpoints
  extractions/              # Analysis run history

Should you commit .alucify/? Yes — it's useful context for your team. The artifacts/ subdirectory contains your specs, and the context files give AI agents structured project context.

Distribution Options

Alucify is available as a VS Code extension, a desktop app, a standalone CLI, and a lightweight MCP server. Choose based on your role:

• Analysis author — Use the Extension, Desktop App, or CLI. Run alucify enable for full Claude Code integration with hooks and commit tracking.
• Consumer developer — Use the MCP Server. A teammate commits .alucify/ and you get context-aware Claude Code with zero setup. Install via alucify mcp-install --with-rules, npx @alucify/mcp, or clone a repo that already has it configured.
• CI / automation — Use the CLI headlessly (alucify analyze-code, alucify analyze-specs, alucify analyze-progress).

See the CLI and MCP Server sections below for details.

Configuration

Settings available in VS Code Settings UI (Cmd+, / Ctrl+,, search "Alucify"):

Commands

Open the command palette (Cmd+Shift+P / Ctrl+Shift+P) and type "Alucify" to see all available commands:

CLI

Alucify ships a CLI for headless use in CI pipelines or terminal workflows.

alucify analyze-code                    # Index codebase (free)
alucify add-spec requirements.pdf       # Add a spec
alucify analyze-specs                   # Analyze specs vs code
alucify analyze-progress                # Analyze changes against requirements
alucify enable                          # Install Claude Code hooks + MCP + CLAUDE.md
alucify disable                         # Remove Claude Code integration
alucify mcp-install [--with-rules]      # Vendor MCP server into .alucify/
alucify plan                            # Plan implementation with Claude Code
alucify continue                        # Close gaps from analyze-progress
alucify capture-intent                  # Extract intent from Claude Code sessions
alucify status                          # View implementation status
alucify context <query>                 # Query project context
alucify usage                           # View token usage summary
alucify check                           # Pre-commit impact check

MCP Server

For developers who just want context-aware Claude Code without the full CLI or extension.

{ "mcpServers": { "alucify": { "command": "npx", "args": ["@alucify/mcp"] } } }

Or install via the CLI: alucify mcp-install --with-rules — this vendors a self-contained server into .alucify/mcp-server.js that can be committed for zero-install team sharing.

License

MIT