# Straion

> Straion is the governance layer for AI coding agents (Claude Code, Cursor, GitHub Copilot). It stores a team's engineering standards as structured, version-controlled rules, selects only the rules relevant to each task, validates the agent's plan before any code is written, and re-checks the resulting changes.

Straion keeps AI-generated code aligned with your architecture, security, and compliance standards, from the first prompt to merge. Import your existing AGENTS.md, CLAUDE.md, and .cursorrules files once, then manage everything from one central hub instead of scattered files that drift across repos. Rules are scoped by team, repository, language, and framework, so each task gets only the guidance that applies. It works alongside your AI coding tools; it governs, it does not generate.

## Product
- [Product overview](https://straion.com/product): What Straion does and how it governs AI coding agents.
- [How it works](https://straion.com/how-it-works): Rule selection by context, plan validation before code, and pull request enforcement.
- [AI coding guidance](https://straion.com/product/ai-coding-guidance): Guide Claude Code, Cursor, and Copilot with company-specific rules so AI-generated code follows your architecture, security, and quality standards.
- [Code review](https://straion.com/product/code-review): Automated pull request validation against non-functional requirements, so code stays aligned with company standards without manual review.
- [Requirements management](https://straion.com/product/feedback-on-requirements): Capture and manage your organization's requirements and keep specifications and code aligned at every step.
- [Pricing](https://straion.com/pricing): Plans and tiers (free tier to start, paid and enterprise tiers for teams).
- [Security and trust](https://straion.com/security): How Straion handles code and data. The CLI runs locally, source code is never stored, rules stay in your own Git, EU hosting, and no model training on your data.

## Solutions
- [Embedded software (MISRA)](https://straion.com/solutions/embedded-software): AI coding agents in MISRA C and C++ codebases. The applicable rules reach the agent before it plans, and every deviation carries a rule, a rationale, an approver and a date. Sits upstream of static analyzers such as Axivion, Polyspace, LDRA, Coverity and Parasoft, and carries no tool qualification of its own.

## Comparisons
- [AGENTS.md vs Straion](https://straion.com/product/straion-vs-agents-md): Why a static markdown file breaks down at team scale, with a before and after import example, and what Straion adds on top.
- [Straion vs custom skills](https://straion.com/product/straion-vs-skills): How Straion compares to custom skills in Claude Code and Cursor.

## Blog
- [You Shouldn't Be Finding Rule Violations in Code Review](https://straion.com/blog/rule-violations-code-review): Straion turns every rule that applies to a change into a decision the agent has to resolve before the work is done, and every exception into one you approve.
- [Straion Rules Now Live in Git](https://straion.com/blog/straion-rules-now-live-in-git): Straion rules can now be managed in Git: versioning, pull request reviews, and full history.
- [Developers Asked Anthropic to Build This. We Did.](https://straion.com/blog/developers-asked-anthropic-to-build-this): A developer filed GitHub issue #14467 on claude-code describing exactly the problem Straion solves.
- [The Real Cost of Building AI Coding Governance](https://straion.com/blog/build-vs-buy-ai-coding-governance): Building your own AI coding governance layer sounds like a config project.
- [90% Rule Adherence](https://straion.com/blog/90-percent-rule-adherence-straion-coding-agents): When Claude Code runs with Straion, rule adherence climbs into the 90%+ range.
- [1M Tokens Won't Save Your Engineering Standards](https://straion.com/blog/1m-tokens-wont-save-your-engineering-standards): Sourcegraph data from 1,281 agent runs confirms the research: more context degrades AI coding performance.
- [Your AGENTS.md is costing more than you think](https://straion.com/blog/delete-your-claude-md-science-says-so): ETH Zurich research: LLM-generated AGENTS.md and CLAUDE.md files cut resolve rates and raise cost.
- [We raised €1.1M to fix AI coding drift](https://straion.com/blog/we-raised-1-1m-to-fix-ai-coding-drift): Straion announces its €1.1M seed round led by Marathon VC and outlines why autonomous engineering needs a rules layer to prevent AI coding drift.

## Company
- [About Straion](https://straion.com/about-us): The team and why Straion exists.
- [Blog](https://straion.com/blog): Research and writing on AI coding governance.