Stop reopening settled debates
Find the original context and trade-offs before repeating an expensive discussion.
ADR Guard — Architecture Decision Records & Governance
Decisions that stay understandable
Architecture decisions deserve more than forgotten documents.
Document, validate, review, and evolve architecture decisions with confidence. ADR Guard connects learning, team workflow, and verifiable automation.
ADR-0042 · Proposed
Adopt Redis for distributed caching
- Context
- Repeated reads increase latency.
- Decision
- Cache-aside with explicit TTLs.
- Consequence
- Lower load; new operational complexity.
Why ADRs matter
Give decisions a durable memory.
Code shows what a system does. An ADR preserves why a consequential choice was made, which alternatives were considered, and what the team accepted in return.
Onboard with context
Help new teammates understand constraints instead of reverse-engineering intent from code.
Trace change over time
Supersede decisions explicitly while retaining the path that led to the current architecture.
What ADR Guard does
A practical guardrail, not an approval robot.
ADR Guard works locally and in delivery workflows. It validates document contracts and relationships; people remain accountable for architectural judgment and acceptance.
Create
Generate canonical Minimal or Extended ADRs, or start from a constrained custom template.
Validate
Check structure, statuses, relationships, changed files, and diagnostic baselines.
Review
Request bounded AI-assisted review with explicit providers, security controls, and human judgment.
Track
Generate an index and keep supersession and relationship history navigable.
Integrate
Use the CLI, official GitHub Action, container, and workspace-hosted VS Code extension.
Draft
Use opt-in AI assistance without treating generated text as an accepted decision.
Choose your path
Start where you are.
Learn the concept, write your first record, design a team practice, or go straight to the toolchain.
Agent Skills
Bring ADR Guard into your agent workflow.
11 portable, task-focused Agent Skills help you discover, document, validate, review and evolve architecture decisions—without giving AI the authority to approve them.
Find the right workflow
Explore six essential and five advanced skills, from when to record an ADR to audits and supersession.
Install for your agent
Use the Skills CLI to select only the skills you need. ADR Guard’s .NET CLI is installed separately.
Try real scenarios
Start with a decision question, compare alternatives, propose an ADR, or review its lifecycle safely.
Decision lifecycle
Keep history. Change direction deliberately.
The canonical states describe the record. Your organization decides who proposes, reviews, and accepts it.
- 1Proposed
The decision is open for analysis and review.
- 2Accepted
Authorized people agreed to follow the decision.
- 3Deprecated
The decision is no longer recommended, but history remains.
- 4Superseded
A new ADR explicitly replaces this decision.
Examples library
See real trade-offs, not empty templates.
Each scenario explains context, alternatives, criteria, consequences, and when its level of detail is useful.
Performance
Distributed caching with Redis
Compare the same decision in Minimal, Extended, and MADR 4.0.
Read example →Integration
Sync or async communication?
Balance coupling, resilience, latency, consistency, and operating cost.
Read example →Evolution
Superseding a decision
Change direction without rewriting the architectural record.
Read example →Make the next architecture decision explainable.
Start with one real decision. Keep the record proportionate, review it with the people affected, and automate only the checks a tool can honestly perform.