Demo
AI tools generate working code quickly. They do not reliably preserve architecture boundaries, public entrypoints, dependency contracts, or architecture decisions.
ArchPilot makes architecture explicit and continuously verifiable in VS Code, CLI, and CI.
What ArchPilot validates
Local-first. Deterministic. No AI required.
Walkthrough
The demo flow is intentionally local. ArchPilot works from repository metadata and deterministic rules so developers can inspect what changed before relying on it in review.
Smart Init
The Smart Init flow inspects repository structure, proposes architecture metadata, and gives developers a concrete starting point instead of a blank config file.
Validation
ArchPilot validates architecture rules locally. The same rule model can run from VS Code, the CLI, and CI without requiring AI or hosted code scanning.
Findings
Findings point to the rule, affected architecture surface, and next action so engineers can decide whether to fix code, update a contract, or document a decision.
Example output
These are examples of the kind of signals ArchPilot surfaces. They are architecture validation findings, not runtime, vulnerability, or security scan claims.
$ archpilot validate
Architecture validation completed
Score: 84/100
high AP-DEP-004 undeclared dependency
billing -> risk/rules
medium AP-DEP-005 non-public import
web imports payments/internal/repository
medium AP-ADR-004 ADR coverage missing
tenant isolation configured without accepted ADRbilling imports risk/rules without a declared dependency in the module contract.
web imports payments/internal/repository instead of the payments public entrypoint.
tenant isolation is configured, but no accepted ADR explains the decision.
Validation score moved from 91 to 84 after new dependency and documentation findings.
Workflow
ArchPilot is designed as a repeatable architecture validation workflow, not a chat prompt or a hosted scanner.
Start from an existing repo and generate reviewable architecture metadata instead of hand-writing the whole model.
Run deterministic checks against contracts, public entrypoints, dependency rules, and architecture docs.
Use local tooling to fill in architecture artifacts where the repository has enough structure to infer them.
Keep architecture feedback near the files engineers are editing, before drift reaches review.
The local validation workflow can run in terminals and pull request automation without hosted scanning.
Use CI results and optional Cloud metadata to make architecture validation results visible during review.
Real repositories
ArchPilot has been exercised against real-world codebases including ArchPilot itself, NestJS services, Next.js applications, Spring Boot projects, Laravel applications, and multi-component repositories.
The goal is practical architecture validation on existing codebases, not toy examples.
How this is different
It is most useful when teams already care about code quality and now need architecture boundaries to stay visible as AI-assisted changes accelerate.
Excellent for import rules. ArchPilot adds architecture contracts, findings, generated metadata, and governance workflow around those boundaries.
Strong Java architecture testing. ArchPilot is multi-stack and shaped around editor, CLI, CI, and optional team visibility.
Great workspace and build tooling. ArchPilot focuses on architecture validation, module contracts, quality, and drift.
Broad code quality analysis. ArchPilot is specifically about architecture boundary governance and deterministic validation.
CI and PR review
Once the local loop is useful, teams can run the same validation in CI and optionally publish governance metadata for shared visibility in ArchPilot Cloud.
Install and try
Install the VS Code extension or try the CLI in an existing repository. Cloud can come later when your team wants shared architecture visibility.
Get started in minutes.
npm install -g @archpilotlabs/archpilot
archpilot --version