Architectural drift prevention for the agentic AI SDLC

Architecturethat holds.

AI writes the code. Your architecture still decides what ships.

Install Mneme Run the 2-minute demo or read the source on GitHub →

Supported by Google for Startups Microsoft for Startups MIT licensed · open source

Two people arranging a wall of hand-drawn application screens into one system — Architecture is decided by people. Mneme carries that decision to every agent.

One decision · one action · one defensible verdict

The code can work and still violate the system.

ADR-017Approved · Platform team

Keep data access behind the service boundary.

Frontend surfaces may request analytics through the backend service. They must not instantiate data-platform clients directly.

IF path = frontend/**
AND dependency = @google-cloud/bigquery
THEN BLOCK

decision owned · reviewed

Agent proposal

// frontend/checkout.ts
import { BigQuery }
  from '@google-cloud/bigquery';

export async function
trackCheckout(event) {
  const bq = new BigQuery();
  return bq.insert(event);
}

Task context: add checkout analytics
Changed surface: frontend route

Mneme verdict

BLOCKED

ADR-017 forbids a data-platform client in the frontend checkout service.

Had it shipped Warehouse credentials bundled into the browser, and a customer-data path nobody reviewed.

REMEDIATION
Call the approved backend analytics service instead.

SOURCE
ADR-017 · Data access boundary

See the enforcement loop

Start with proof. Then use a real boundary.

Recommended · 2-minute proof

Watch Mneme catch architectural drift.

Follow one proposed change from an approved decision to a deterministic verdict. No API key or project setup required.

01Propose a changeAn agent requests a path the architecture forbids.
02Match the decisionMneme resolves the governing ADR and constraint.
03Return evidenceThe verdict names the source and permitted next action.

WARN · ADR-005Traceable · reproducible

Walk the full demo

When the proof is clear

Test one boundary from your system.

The quickstart walks through the smallest useful evaluation: one approved decision, one proposed change, and one verdict your team can inspect.

DecisionOne current ADR
ChangeOne representative diff
EvidenceOne sourced verdict

Read the quickstart

For teams

Architecture is a human decision. Enforcement should preserve that authority.

A Mneme pilot turns one real set of architectural decisions into an enforceable workflow across the agents and repositories your team already uses. Start with one boundary, one repository, and evidence everyone can inspect.

People decide. Mneme makes the decision operational. Evidence proves it held.

Request a pilot See the team rollout path

Frequently asked

Three direct answers.

What is Mneme HQ?

Mneme is open-source architectural drift prevention for the agentic AI SDLC. It turns approved ADRs, engineering standards, project context, and prior decisions into deterministic guide, warn, or block verdicts before incompatible changes land, with every verdict traced to its source.

When does Mneme enforce architectural decisions?

Mneme can apply relevant decisions before an agent writes code, during file changes through supported hooks, and in CI before a change merges. Each guide, warn, or block verdict identifies the governing source and permitted next action.

How is Mneme different from context, auditing, and code review tools?

Context systems help agents understand. Auditing tools show how agents behaved. Code review evaluates completed output. Mneme enforces the approved engineering intent the agent must follow, while linking every verdict to its source.

Explore

A table covered in written decisions, one held in the hand and picked out in colour

One clear home for every layer of the story.

ProductHow Mneme works ArchitectureDecision and enforcement pipeline IntegrationsAuthoritative support matrix ComparisonsWhere each tool belongs DemosRunnable product evidence EvidenceBenchmark and methodology

Verified compatibility

One decision corpus. Honest support levels.

View the authoritative integration matrix →

Native: Mneme CLI
Claude Code hook
Claude Agent SDK hooks
Google Antigravity adapter
Codex CLI adapter
Kiro hook (CLI 3.0)
Tested workflow: Cursor rules export
GitHub Actions gate
GitLab CI gate
Paperclip (via Claude Code)
Designed to support: GitHub Copilot · OpenCode
JetBrains · VS Code · Warp
In progress: Deep Agents middleware POC

Works with Claude Code Agent SDK Cursor GitHub Actions GitHub Copilot Windsurf OpenAI Aider Google Antigravity + more →

Architecture that holds

Give every coding agent the decisions your team already made.

Request a pilot Run the 2-minute demo