@reality-architect/cli

Unified Cockpit CLI & Stdio MCP Server for Reality Architect — context grounding, reality.md conformance, and empirical diffusion assessment for AI agent harnesses.

Part of the Reality Architect ecosystem.

Capabilities

• reality.md Conformance Engine: Check any reality.md against formal L0–L3 conformance levels.
• Harness Projection Emitter: Compile tailored context projections for Claude, Cursor, OpenCode, Codex, and Antigravity.
• Empirical Diffusion Engine: Assess agent claims, code blocks, or factual statements against 6 empirical indicators, detect failure modes, and apply certainty caps.
• Stdio Model Context Protocol (MCP) Server: Exposes 5 production tools directly to LLM agent harnesses.

Installation & CLI Usage

Run without installing:

npx @reality-architect/cli --help

Or install globally:

npm install -g @reality-architect/cli

Commands

# Validate conformance
reality-architect validate ~/reality.md

# Emit Claude context
reality-architect emit ~/reality.md --target claude

# Output dependency graph
reality-architect graph ~/reality.md

# Assess empirical claim confidence & check certainty capping
reality-architect assess "The API endpoint has 100% guaranteed zero downtime"

# Launch Stdio MCP Server
reality-architect --mcp

Model Context Protocol (MCP) Integration

Add to your claude_desktop_config.json or Antigravity MCP settings:

{
  "mcpServers": {
    "reality-architect": {
      "command": "npx",
      "args": ["-y", "@reality-architect/cli", "--mcp"]
    }
  }
}

Exposed MCP Tools

1. reality_validate: Validates a reality.md file or text against conformance levels (L0-L3).
2. reality_emit: Emits tailored context files (claude, cursor, opencode, codex, antigravity).
3. reality_brief: Extracts the next-artifact build gap brief.
4. reality_graph: Outputs the node and edge invariant dependency graph.
5. reality_assess: Evaluates empirical reality confidence and applies certainty caps to prevent hallucinations.

License

MIT © Frank Riemer / Reality Architect