Skip to main content
TRW

Raise the unit of accountability from one team to the whole organization

Claude tells you how much your teams spent on AI. TRW tells you what they actually built with it, whether it was verified against its requirements, and whether it will still make sense next quarter — a record that travels with the work, not a dashboard of who-spent-what. And it runs where you need it — on the developer’s machine, in your own cloud, or fully air-gapped.

Agents proliferated. Org-level accountability did not.

AI coding agents now write across every repo, in every team, on every host. Spend is metered. Access is logged. The work itself — what got built, whether it was verified, whether the knowledge survives — has no record above the level of a single engineer’s session.

Spendlogged
Workmissing

every dollar is metered — the work itself goes unrecorded

Spend is measured. Work is not.

You can see seats consumed and tokens billed. You cannot see which requirement an agent built against, or whether a build-check verdict passed before it shipped. The vendor consoles bind a seat to a cost — never a requirement to a verification verdict.

One team is not the organization

A single-team console shows one team’s runs in isolation. It cannot answer the org question: across every team and every host, what was built, what was verified, and where the same problem got re-solved from zero.

Knowledge that evaporates by team

Each team’s discoveries die at the end of a session, or sit in a notes file no other team will ever read. One team’s verified work does nothing to raise the floor for the next. The organization re-pays for the same lesson, team by team.

Governance across time, not governance of access

Every memory and governance product in this space governs access — who used which model, on which seat, at what cost — and is locked to one host to do it. TRW governs work across time: it binds a requirement to whether its build-check verdict passed, across sessions, in a record that travels with the work. It sits below the GRC attesters and above the generators — the layer that records the work so the platforms above it can attest to it.

Access tells you the bill. Work tells you what you bought.
How host-locked access governance compares with TRW work governance
DimensionHost-locked access governanceTRW work governance
Unit of analysisUsers, seats, spend — access metered per hostRequirement bound to a build-check verdict — work recorded across sessions
What the record provesThat AI was used, and how much it costWhat was built, whether it was verified against its requirements, whether it still holds
Vendor couplingLocked to one IDE, model, or vendor consoleAbove any IDE, model, or host via MCP — governance is not hostage to one vendor
Where it runsVendor cloud, on their termsYour infrastructure — local, your cloud, or air-gapped

What ships today for an organization

Every capability below is in the product now. This is the honest evaluation surface for a security and procurement review.

SHIPPING

Org & roles

Team workspace

Single-org team workspace — invite teammates by email with a role, create sub-teams, add and remove members

Roles

Owner / admin / member / viewer, with cross-tenant privilege-escalation guards

Data model

Org / team / membership entities; architecture designed for N-level org hierarchy

Analytics

Org-scoped analytics summary

Evidence & data rights

Audit trail

Persisted, org-scoped audit trail of org and membership activity

Content hashing

SHA-256 content hash on every record

Data subject rights

GDPR JSON export and erasure

Identity & deployment

Identity & keys

Per-org API keys; JWT / OAuth / 2FA authentication

Open framework

trw-mcp and trw-memory run repo-local, telemetry off by default — the work stays on the developer’s machine

Runs where you do

On the developer’s machine, in your own cloud, or fully air-gapped

Vendor review surface

OWASP-hardened, source-available codebase under BSL-1.1 — readable, auditable, license-clear

Run TRW in your environment

Enterprises want the evidence layer inside their own perimeter. TRW is built for that. The open framework runs entirely on your infrastructure today; the platform can be deployed in your cloud or on-prem through the engagement.

YOUR PERIMETER

Air-gapped

AVAILABLE NOW

Open framework, installed with no network. Zero outbound calls.

On-prem / your cloud

AVAILABLE NOW

Open framework on your own AWS, VPC, or on-prem host. Repo-local memory.

Platform in your cloud or on-prem (BYOC)

ENTERPRISE

The hosted platform / control-plane deployed inside your environment. We deploy with you.

TRW-managed SaaS

OPT-IN

The hosted platform, run by TRW. Opt-in; org-scoped audit recorded server-side.

Built to be evaluated by your security team

A small, honest surface to assess — source-available where it can be, conservative where data leaves the machine, and explicit about exactly what is enforced and how.

Local-first open framework

The open framework — trw-mcp and trw-memory — runs repo-local with telemetry off by default. A default install makes zero outbound calls, so it runs on your infrastructure today: a laptop, your AWS or VPC, on-prem, or fully air-gapped. The work and the memory stay where you put them.

Source-available, OWASP-hardened

trw-mcp and trw-memory ship under BSL-1.1 — source-available and readable end to end, so your security team can read the code they are evaluating. The open packages present a clean, OWASP-hardened surface for a standard OSS vendor review.

Credential hygiene and identity

Per-org API keys and JWT / OAuth / 2FA authentication. Roles carry cross-tenant privilege-escalation guards on every boundary. Secrets are not logged.

Isolation and record integrity

Tenant isolation is enforced at the application layer, with cross-tenant privilege-escalation guards on every role boundary. Every record carries a SHA-256 content hash. We document the integrity model in plain terms so your reviewers can verify exactly what it guarantees.

The evidence source attesters attest against

TRW is the engineering-side record your GRC platform attests against — readiness, not a certification. We map capabilities to control families with “aligns with” and “maps to,” never “certified.” Not legal advice.

Generators
Claude Code, Cursor, Copilot
Produce the work
TRW
The engineering-side record
Records what was built and whether it was verified
GRC attesters
Credo AI, Modulos
Attest to it

The stack: generators on top — Claude Code, Cursor, Copilot — produce the work. GRC attesters above — Credo AI, Modulos — attest to it. TRW sits between: below the attesters and above the generators, the layer that records what was built and whether it was verified, so the platforms above it have something true to attest to. TRW does not replace your GRC platform; it gives it evidence.

EU AI Act

Maps to

Aligns with audit-trail and transparency readiness. Whether AI-assisted coding is high-risk is an open question we do not assert.

NIST AI RMF

Maps to

Maps to Map / Measure / Manage practices via phase-gated runs, requirements traceability, and a persisted audit trail — a readiness signal, not a certification.

ISO/IEC 42001

Maps to

Aligns with the AI-management-system evidence expectations — run governance, role ownership, documented reversion. TRW supplies the underlying record, not the certificate.

Proof by practice, not by logo

TRW is built with TRW. The post-generation lifecycle — the team product, the role model, the audit trail, and the memory layer — is exercised every day on this monorepo. These are tracked repo metrics from that work, not effectiveness claims.

Sprints

882+

recorded internal sprint cycles on this monorepo

PRDs

2,862+

requirements tracked through phase-gated runs

Tests

20,266+

tests present across the codebase

Learnings

3,469+

active learnings captured and recalled across sessions

No customer logos and no outcome-lift numbers — just the record of building TRW with TRW. If you want to evaluate the mechanism on your own org, that is what the design-partner program is for.

Questions a security and procurement review will ask

requirement → verified → sha256:9f2c…a17eSEALED

Govern the work, not the spend — across your whole organization

A requirement bound to a verification verdict, knowledge that compounds and is governed, above any host via MCP, running on your own infrastructure, and ready for the evidence your attesters attest against. If your organization is running AI coding agents across multiple teams, the next step is a conversation about rollout fit.

Keep going

If you haven't installed yet, start with the quickstart. If you have, the lifecycle page explains what TRW actually does once it's running.