# TRW — Accumulated Intelligence for AI Agents

> The methodology layer for AI-assisted development. Every AI coding tool resets to zero. TRW is the one that doesn't.

## What TRW Is

TRW (The Real Work) is the engineering methodology layer that sits above AI coding tools — Cursor, Devin, Claude Code, Copilot, Codex, Aider — and makes them compound. It is not another coding tool. It is the discipline of context engineering that makes coding tools produce reliable, traceable, improving results.

Five integrated pillars that no competitor combines:

1. **Knowledge Compounding** — Learnings persist across sessions with Q-learning scoring and Ebbinghaus decay curves. High-impact discoveries auto-promote into permanent project context. Session 50 is measurably better than session 1.
2. **Requirements Engineering** — AARE-F framework with 12-dimension validation, content density scoring, and FR-by-FR traceability from spec to implementation to verification. The spec-driven development (SDD) approach that turns vibe coding into verified engineering.
3. **Sprint Orchestration** — Parallel agent teams with file ownership contracts, structured handoffs, and delivery gates. 64+ sprints executed through the framework.
4. **Phase Gates** — 6-phase lifecycle (Research → Plan → Implement → Validate → Review → Deliver) with build checks, adversarial audits, and independent verification.
5. **Agent Teams** — Multi-agent coordination with lead/implementer/tester/reviewer/auditor roles, YAML interface contracts, and file ownership boundaries.

## Installation

```bash
curl -fsSL https://trwframework.com/install.sh | bash
```

Or add directly to your MCP config (~/.claude/mcp.json):

```json
{
  "mcpServers": {
    "trw": {
      "command": "uvx",
      "args": ["trw-mcp"]
    }
  }
}
```

Works with Claude Code, Cursor, OpenCode, Codex, and Aider via client profiles.

## Architecture

Four packages in a monorepo:

- **trw-mcp** (v0.40.0) — Python MCP server (FastMCP, Pydantic v2): 24 tools, 6 resources, 24 skills, 12 agents. Available on PyPI under BSL 1.1.
- **trw-memory** (v0.6.6) — Standalone memory engine (SQLite, sqlite-vec): hybrid retrieval (BM25 + dense vectors), knowledge graph, lifecycle management. Available on PyPI under BSL 1.1.
- **backend** — Platform API (FastAPI, SQLAlchemy): 22 routers, JWT auth, Procrastinate workers, intelligence pipeline
- **platform** — Frontend (Next.js 15, TypeScript, Tailwind CSS): admin dashboard, marketing site, documentation

## Proof

TRW was built by AI agents using TRW — every PRD, every sprint, every learning is real output from the system governing its own development:

- 64+ sprints executed through the framework
- 225+ PRDs through the full AARE-F lifecycle
- 8,000+ tests across all packages
- 91% test coverage on the memory engine
- 654+ active learnings accumulated from dogfooding
- 5 client profiles for cross-tool compatibility
- Language-agnostic: works with Python, TypeScript, Go, Rust — any language

## How It Works

Every task flows through six phases:

1. **RESEARCH** — Load prior learnings, audit codebase, register findings
2. **PLAN** — Design approach, identify dependencies, create execution plan
3. **IMPLEMENT** — Execute with periodic checkpoints that preserve progress across context compaction
4. **VALIDATE** — Run tests + type-check, verify coverage meets threshold
5. **REVIEW** — Independent quality audit, fix gaps, record discoveries
6. **DELIVER** — Sync artifacts, promote high-impact learnings, close run

## Key Differentiators

- **Offline-first**: Everything runs locally in a `.trw/` directory. No cloud required for core functionality.
- **Zero lock-in**: One config entry to add, one config entry to remove. Your code stays untouched.
- **Self-improving**: The learning system uses the same scoring and decay curves it provides to users.
- **Verified claims only**: No feature that isn't tested. No metric that isn't measured.
- **Tool-agnostic**: Works with any MCP-compatible AI coding tool via client profiles.
- **Context engineering**: Goes beyond prompt engineering to manage the full lifecycle of AI agent knowledge.

## Competitive Position

TRW is the only framework that combines all five pillars — knowledge compounding, requirements engineering, sprint orchestration, phase gates, and agent teams — into a single integrated system. Verified across 80+ competitor analysis. Closest alternatives:

- **Kiro (AWS)** — SDD workflow but no knowledge compounding
- **BMAD** — Multi-agent agile but stateless across sessions
- **Mem0** — Memory storage but no scoring, decay, or promotion
- **CrewAI/LangGraph** — Agent orchestration but no requirements lifecycle or learning
- **GitHub Spec Kit** — Requirements templates but no lifecycle management

## Pages

- **Homepage** (/) — Overview, interactive demos, growth visualization, live metrics, FAQ
- **Memory** (/memory) — Interactive explanation of the memory system: tiers, search, knowledge graph, decay curves
- **For Developers** (/for/developers) — Individual developer value proposition and workflow demo
- **For Teams** (/for/teams) — Team governance, quality process, and coordination capabilities
- **Documentation** (/docs) — Getting started, tools reference, skills, agents, lifecycle, API, configuration
- **Quickstart** (/docs/quickstart) — Two-minute installation and first session guide
- **About** (/about) — Origin story, what TRW became, where it's going
- **Pricing** (/pricing) — Free during private beta, team plans coming
- **Waitlist** (/waitlist) — Early access signup
- **Metrics** (/metrics) — Live framework health metrics from dogfooding
- **Contact** (/contact) — Contact information
- **Privacy** (/privacy) — Privacy policy (offline-first, opt-in telemetry)
- **License** (/license) — BSL 1.1 software license terms

## Links

- Website: https://trwframework.com
- Documentation: https://trwframework.com/docs
- Memory System: https://trwframework.com/memory
- For Developers: https://trwframework.com/for/developers
- For Teams: https://trwframework.com/for/teams
- Early Access: https://trwframework.com/waitlist
- Detailed Info: https://trwframework.com/llms-full.txt