What is LingXi

LingXi is a Cursor plugin that gives developers a persistent memory workflow — making AI work the way you do.

The Problem

When working with AI coding tools like Cursor, you likely run into these issues:

• Every conversation starts from scratch — AI doesn't remember your past decisions, preferences, or lessons learned
• No engineering process — AI jumps straight to code without requirements analysis, design review, or code review
• Context overload — As conversations grow longer, information gets noisy and AI output quality drops
• Team knowledge stays siloed — Each person's AI collaboration experience is locked inside their own chat history

How LingXi Solves This

🧠 Persistent Memory

LingXi captures your judgments, preferences, and lessons learned during development through three capture paths: automatic (automatic session distillation), manual (/remember, /init), and workflow taste sniffing (collecting your choices during task/plan/build/review when context calls for it; see Taste Sniffing), distilling them into structured "memory notes." In every new conversation, LingXi automatically retrieves and injects the most relevant memories, so AI truly "knows" how you work.

🔄 Self-iterate

LingXi self-iterates: it continuously collects system run state and, on a default 24-hour heartbeat, audits that state. Low-risk improvements found during audit (e.g. memory merge suggestions, index completion) are applied in the background—no extra action from you, and your main conversation is never interrupted; you only get a short brief. Self-iterate is a global feature and will gradually extend to the workflow, gating, and more subsystems, so LingXi becomes more stable and more attuned to you the longer you use it.

🔄 Flexible Workflow

An end-to-end development flow driven by task, vet, plan, build, review Skills:

task → vet → plan → build → review

Invoke them explicitly via /task, /plan, etc. or natural language (e.g. “run task”). These workflow Skills are for manual or explicit invocation only; they are not auto-loaded by semantic match. The flow is composable on demand with decoupled entry points; you decide when to use each skill.

When to use this workflow: The LingXi workflow is intended for tasks that are larger than ~3 person-days and of medium-to-high complexity — such tasks benefit from upfront architecture and solution design; task + vet help architects (or whoever owns the design) shape the architecture and overall plan. For simpler tasks, using your IDE's Agent mode is enough; there is no need to start the LingXi workflow.

🛡️ Human in the Loop

AI never acts on its own. Key decisions — requirement confirmation, memory writes, approach selection — always need your approval.

Design Philosophy

Principle	Meaning
In Sync With You	Persistent memory so the AI works the way you do
AI Native	Respect AI capability and leave room for evolution; key decisions are human-led, with gates
To Your Liking	Lower cognitive load, smooth user-friendly experience

How This Differs From Cursor Rules and Other Approaches

LingXi provides persistent memory plus a structured workflow, unlike static Cursor Rules or one-off prompts: it focuses on cross-session "learning" and a decoupled, on-demand workflow (task / vet / plan / build / review Skills). Memories are captured through three paths — automatic (automatic session distillation), manual (/remember, /init), and workflow taste sniffing (context-driven during task/plan/build/review; see Taste Sniffing) — and injected when relevant in new conversations; self-iterate is a global feature that runs diagnosis and auto-improvements in the background on a schedule, with memory-related improvements implemented today and more of the system to be covered over time.

Next Steps

Ready to get started? Head to Quick Start to install LingXi in your project.