mixai-memory

A set of rules for keeping persistent project memory with Mixai. It tells a coding assistant how to search past decisions, save new ones, and reload project context.

In plain words
What is it for?
Use it to search or save project memory, load a context brief, and prepare a handoff to another coding assistant when needed.
Why use it?
It reduces repeated repository exploration and helps preserve decisions after the conversation is shortened or restarted.

Cursor rule for Cursor

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add rules/thinhledev/mixai/mixai-memory
Clone the repo
git clone --depth 1 https://github.com/thinhledev/mixai

Made for: Cursor.

Per session 268 This file is loaded in full into every session.
When invoked 268 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00268 $0.00268
Opus 5 $0.00134 $0.00134
Sonnet 5 $0.00054 $0.00054
Haiku 4.5 $0.00027 $0.00027

Measured yesterday against content hash adf322e0ceb8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mixai-memory scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.cursor/rules/mixai-memory.mdc · 11 lines

What it actually says

Mixai project memory

This project has persistent local memory managed by mixai. Before re-exploring the repository from scratch, check it first: use the memory_search MCP tool or mixai memory search "<topic>" for past decisions, constraints, and outcomes (results are compact previews — drill in with memory_get or mixai memory show <id>). Record new durable decisions the moment they are made with the memory_save MCP tool. If your context feels thin after a compaction or reset, run mixai brief (or the context_brief MCP tool) to reload the latest objective, live workspace state, and the most relevant project memory without re-reading the repository. If you or the user hit a usage or rate limit, or the user asks to switch coding agents, call list_handoff_targets to see which other agents (Codex, Claude Code, Cursor, Antigravity) are installed and ready right now, then — only after the user confirms — call switch_to_provider to hand off: it checkpoints this session's work and launches the target agent with a continuation prompt.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 11 lines · 268 tokens per session scan A adf322e0ceb8

Subscribe to this mod's changes

mixai-memory is a cursor rule published in the GitHub repository thinhledev/mixai (2 stars, last pushed 17d ago), licensed MIT. It adds 268 tokens to every session, about $0.0013 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.