.cg/ - the memory your AI agents don't have
Every new AI session starts from zero. Yours doesn't have to. cxgrd builds a .cg/ folder, your codebase's memory
No re-explaining your architecture to the agent every session.
Tell your AI what exactly will break before it even makes a change.
cxgrd gives your AI coding tools — Cursor/Claude Code/Windsurf — a map of your codebase, so they can avoid breaking something three files away.
See it in action
See CXGRD in action with this quick walkthrough.
Not another agent guardrail
Cursor, Claude Code and other tools guard what the AI does in your editor — permissions, hooks, rules, one session at a time. CXGRD guards what actually merges — dependency-graph-aware, enforced org-wide, regardless of which tool or person wrote the code.
Why teams are interested
AI coding tools are fast, but they are often stateless and context-limited. CXGRD adds architectural memory and dependency-aware safety checks so velocity does not come at the cost of production stability.
See what breaks before you prompt
Most review tools tell you what broke after the AI already wrote the diff. cxgrd shows you what will break before you even send the prompt
Steer the agent instead of cleaning up after it.
Prompt enrichment for coding agents
Generate architecture-aware prompts with the right file context instead of relying on temporary agent memory.
Improves AI output quality on large codebases
A compiler check, not a second AI
Compilers verify every AI-generated diff—not another AI guessing whether the first one got it right.
If it doesn't compile clean, it doesn't ship
How CXGRD works
A command-first flow that fits naturally into modern AI-assisted coding loops.
scan
Crawls the repo and builds dependency + symbol graphs.
Creates and updates .cg/ graph memory.
input
Analyzes a planned change and computes blast radius.
Returns impacted files and architectural dependencies.
prompt
Builds an architecture-aware prompt for your AI assistant.
Produces enriched prompt context for safer code generation.
check
Verifies implementation quality with compiler-backed checks.
Flags structural issues before merge or commit.
Architecture
A lightweight TypeScript CLI paired with a cloud backend — no native dependencies, works anywhere Node runs.
TypeScript CLI
- Dependency graph construction from static import analysis
- Blast radius computation across the full file graph
- Compiler-backed checks via tsc, pyright, and cargo
Cloud backend
- Shared team graph storage and health snapshots
- GitHub App integration for PR commit status enforcement
- Merge policy evaluation, audit log, and team dashboard
See the architectural diagram here.