Introduction
Context engineering for knowledge and decisions.
Sruja helps teams capture important engineering knowledge, retrieve the right task context for developers and AI agents, and verify that changes still align with those decisions. Optional repo.sruja is reviewed intent in Git, not the day-one requirement.
New here? Start with Quick start to learn the core loop: capture, retrieve, verify. Add
repo.srujaonly when you want reviewed intent in Git.
The Problem
How does your AI know the real architecture today?
| Your approach | Problems |
|---|---|
| Raw model context | Easy to miss boundaries, invent dependencies, or forget prior decisions |
| Drawings in Miro/LucidChart | Manual updates, easy to forget, drifts from code |
| Wiki pages | Inconsistent, hard to maintain, no validation |
Sound familiar? You're not alone. Most teams struggle with this.
The Solution
A small context loop with optional reviewed intent.
With Sruja:
- Capture knowledge and decisions with docs, decision records, and optional reviewed intent
- Retrieve grounded context with
focus,ai, and MCP before the host agent edits - Verify the result with drift, intent, and
verify-task - Add reviewed intent in Git (
repo.sruja) only when the team wants stricter governance
Sruja is not a replacement for Cursor or Copilot — it is the guardrail layer beside them.
How This Helps
| Before Sruja | With Sruja |
|---|---|
| AI guesses from partial context | AI works from repo evidence and linked decisions |
| Knowledge lives in stale docs or chat history | Important context is captured and retrievable |
| Hard to catch generated mistakes | Drift, intent, and verification gates catch regressions |
| Hard to brief agents consistently | Task-scoped context is reusable |
| Diagrams become the truth | Diagrams are optional outputs, not the product center |
Start Here
- Quick start: core capture, retrieve, verify loop
- Getting started: core workflow plus optional reviewed intent
- CLI guide: daily commands, CI-friendly outputs, and workflows
- VS Code extension: editor commands, diagnostics, and previews
What Sruja Optimizes For
- Evidence over guesses: context starts from what exists in code and linked docs today
- Small surface area: a few core workflows used consistently
- Explicit trade-offs: reviewed intent is optional until the team needs it
- Host-owned LLM loop: Sruja never replaces your editor or agent
- Knowledge and decisions: context engineering, not a feature catalog
Who is Sruja For?
Software Architects
- Review changes against evidence and intent
- Prevent drift through automated gates
- Document decisions with ADRs (Architecture Decision Records)
- Keep reviewed truth in Git when needed
Developers Using AI
- Brief agents before edits with the right task context
- Reduce risky guesses from incomplete repo understanding
- Verify generated changes before calling work done
- Keep prior decisions visible at change time
Teams Maintaining Shared Knowledge
- Capture docs and decisions in a durable, reviewable way
- Integrate into CI/CD so alignment is checked continuously
- Preserve reasoning so future maintainers and agents do not start cold
- Export artifacts only when communication requires them
Next Steps
- New to Sruja? Start with Quick start
- Need workflows? Read the CLI guide and How Sruja works
- Ready for reviewed intent? Follow Using Sruja in your project