Note: Cannot be usefully used with ChatGPT, which has too many inbuilt prejudices for this kind of tool. Works with all other major LLMs.
Reverse Engineer Logic (rel) applies the forensic rigor of binary reverse engineering to human and institutional messages, emails, negotiations, contracts, policies, dialogues, and strategic situations.
Instead of offering polite summaries or skipping uncomfortable subtext, rel treats every sentence, adversative pivot, word choice, actor incentive, and conspicuous omission as an addressable unit inside a cryptographic Obligation Ledger (obl_...). An analysis is rejected as incomplete until 100.0% of sentences and strategic silences are expounded and sealed in a verifiable Completion Ledger (ecl_...).
Ships as:
- Native macOS Universal App (
Reverse Engineer Logic.app, Apple Silicon + Intel, macOS 13.0+): Standalone macOS app with zero Node.js requirement, Keychain-backed multi-provider LLM support, Draft V1 vs. V2 Comparator, and one-click MCP registration across all detected AI agents. - Unified CLI & MCP Server (
rel/rel mcp): Works across macOS and Linux with the same multi-client architecture as REA. - Just download it and ask your AI to install and use it. It needs no special permissions or anything.
- Anthropic Claude (
ANTHROPIC_API_KEYor localclaudeCLI) - Google Gemini (
GEMINI_API_KEY/GOOGLE_API_KEYor localgeminiCLI) - xAI Grok (
XAI_API_KEY) - DeepSeek (
DEEPSEEK_API_KEY) - Mistral (
MISTRAL_API_KEY) - OpenRouter (
OPENROUTER_API_KEY) - Ollama Local Runtime (
http://127.0.0.1:11434- Llama, Qwen, DeepSeek, Mistral, Gemma, etc.) - Antigravity CLI (
agy) - Built-in Deterministic Pragmatic & Game-Theoretic Engine (
builtin): Instant (<50ms), 100% offline structural, rhetorical, and incentive synthesis with zero API keys.
Configures both the rel stdio MCP server and SKILL.md across:
- Claude Desktop & Claude Code (
.claude-pluginincluded) - Cursor (
.cursor-pluginincluded) - Google Antigravity & Gemini CLI (
gemini-extension.jsonincluded) - Windsurf
- VS Code (MCP) & Cline
- Zed
- OpenCode
- Download
Reverse-Engineer-Logic-1.0.0-macOS-Universal.dmgfrom the Releases page. - Drag
Reverse Engineer Logic.appto/Applications. - Open the app -> click AI Agent Integrations (MCP) -> Configure All Detected Agents (or LLM Providers to add an API key / local Ollama model).
curl -fsSL https://raw.githubusercontent.com/MaxFreedomPollard/reverse-engineer-logic/main/install.sh | bashOr from a local clone:
node bin/rel.mjs setup --all-detected
node bin/rel.mjs doctorFor every single unit (S01..Sn), rel expounds:
- Literal Surface Claim: What the text states at face value.
- Operational Function: What tactical job the unit actually performs (social anesthetic before an adversative "though/but", passive blame erasure "mistakes were made", bad-cop authority shielding "committee/policy", or asymmetric extraction).
- Underlying Logic Chain: Explicit input -> implicit inferential bridge -> intended behavioral effect.
- Unstated Load-Bearing Premises: Smuggled assumptions packaged as common ground.
- Four-Layer Motivational Stack:
- Surface Motive (public optics),
- Tactical Motive (immediate conversational leverage),
- Core Self-Interest (payoff, resource, optionality, or risk transfer),
- Ego / Face Protection (status, moral self-image, plausible deniability).
- Word-Choice & Syntax Forensics: Exact mechanics of hedges, minimizers, pronouns, anchors, and escape clauses.
- Counterfactual Candid Translation: What an unhedged actor with zero hidden agenda would have said instead.
- Competing Motivational Hypotheses (
H1,H2,H3): Probability-weighted competing explanations. - Falsification / Diagnostic Probe: A concrete question or move whose observable reaction separates
H1fromH2.
Plus dedicated dossiers for Conspicuous Omissions (OM01..OMk), an Actor Incentive, Fear & BATNA Matrix, and a Strategic Response Playbook.
# Full logic & motivation investigation (writes report.md, evidence-bundle.json, and logic-xray-workbench.html)
rel analyze "Your message or situation..." --context "Who, stakes, and history" --out ./rel-out
# Instant (<50ms) offline deterministic synthesis
rel analyze --fast "Mistakes were made, though policy prevents retroactive changes right now."
# Compare two drafts or statements to expose walked-back commitments & newly injected hedges
rel compare --left "We will sign the $120k contract on Tuesday." \
--right "We might explore a framework down the road when bandwidth allows."
# Verify cryptographic SHA-256 Evidence IDs (ev_...) & 100% Completion Ledger (ecl_...)
rel verify ./rel-out/evidence-bundle.json
# Inspect providers & configured AI agents
rel doctorMIT License. See LICENSE.