M
mindemory/proxenos
A free, open-source agent that keeps your own records private on your computer and drafts the next step for…
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
- ★ 数取值
- 4 ★ · mindemory/proxenos · 取自 GitHub 搜索 API · (2026-11-13 20:00:50 核对)
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型:开源项目
- 收录与活跃
- 收录于 2026-11-04 · 最近见到 2026-11-14 · 共见到 11 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 4 ★ · 1 个来源 · repo
常见问题(答案都来自我们的记录)
它适合谁 / 用来做什么?
我们把它归在这些标签下: 开源项目。标签来自条目自身,不是我们强加的分类。
谁维护它、还在更新吗?
仓库
mindemory/proxenos;我们收录于 2026-11-04;最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# Proxenos
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
A free, open-source personal memory toolbox. Your records stay on your computer; it drafts the next step for you to send.
macOS 13 or newer on Apple silicon (.dmg), or Windows x64 (.exe). Both are unsigned for now: on a Mac, open System Settings → Privacy & Security → Open Anyway; on Windows, when SmartScreen warns, choose More info → Run anyway.
Proxenos reads your own mail, chats, calendars, notes, documents and health records into a private vault on your computer, keeps a written model of who you are, and drafts the next step in your voice: the reply, the plan, the call script, the brief for your next appointment. It was built first for people getting through the health care system, and now it's for anyone.
## Get started
From source, on a Mac with Homebrew:
bash
git clone https://github.com/mindemory/proxenos && cd proxenos && scripts/install.sh
It installs what's missing (Node, Ollama's own app, and Miniforge if there's no conda), sets up the proxenos conda env and the page, and pulls a local model that fits your Mac's memory. Running it again only adds what's missing. --no-model skips it if you'll use Claude; --start-at-login keeps the page and a sync running from login on.
bash
conda activate proxenos
proxenos demo # try it on a made-up person first; prints how to open it
proxenos web # your own, at http://127.0.0.1:4747
Setup asks your name, where to keep the vault and which model to use, then Connect adds a first source. A model is needed for anything to be drafted: an open one on your computer through Ollama (gemma4:26b fits a 32 GB Mac), or Claude with your own Anthropic API key. Windows and Linux run from source too, though less tried (docs/apps.md). Every step by hand is in the guide; every command, in docs/commands.md.
## How it works
It reads your own records. About 30 sources, each on its own: WhatsApp, Slack, Gmail, Microsoft 365, the Mac's Mail and Calendar, Google Takeout, Apple Health and patient portals, Spotify, Strava, notes, PDFs and other exports. Re-reading one changes nothing. Some readers expect a particular export format (docs/datasources.md).
One model of you, in text you can read. Observers read your side of each conversation, a month at a time, and write notes about you. The master folds them into one self-model of your interests, voice and routines, which you can read and correct. Small agents, "clones", each read it and do one job: tasks, where your care stands, a feed of papers, a map of people, a searchable memory (docs/architecture.md).
It drafts, you send. A reply for each chat waiting on you, written the way you write to that person, and emails, plans, call scripts and filled-in forms. You press send. The two exceptions are yours to start: an outbox you switch on per app, only where its drafts have gone out nearly unchanged many times (docs/memory.md), and a phone call you place, where it says it's automated and never agrees to pay (docs/calls.md).
Your work and your life. For each project or paper it writes where things stand and, given the code, does the next step on a branch for you to accept. It also keeps your job leads, a journal of each day, a diary of your life by month, and your weeks of listening, movement and messages over time.
macOS 13 or newer on Apple silicon (.dmg), or Windows x64 (.exe). Both are unsigned for now: on a Mac, open System Settings → Privacy & Security → Open Anyway; on Windows, when SmartScreen warns, choose More info → Run anyway.
Proxenos reads your own mail, chats, calendars, notes, documents and health records into a private vault on your computer, keeps a written model of who you are, and drafts the next step in your voice: the reply, the plan, the call script, the brief for your next appointment. It was built first for people getting through the health care system, and now it's for anyone.
## Get started
From source, on a Mac with Homebrew:
bash
git clone https://github.com/mindemory/proxenos && cd proxenos && scripts/install.sh
It installs what's missing (Node, Ollama's own app, and Miniforge if there's no conda), sets up the proxenos conda env and the page, and pulls a local model that fits your Mac's memory. Running it again only adds what's missing. --no-model skips it if you'll use Claude; --start-at-login keeps the page and a sync running from login on.
bash
conda activate proxenos
proxenos demo # try it on a made-up person first; prints how to open it
proxenos web # your own, at http://127.0.0.1:4747
Setup asks your name, where to keep the vault and which model to use, then Connect adds a first source. A model is needed for anything to be drafted: an open one on your computer through Ollama (gemma4:26b fits a 32 GB Mac), or Claude with your own Anthropic API key. Windows and Linux run from source too, though less tried (docs/apps.md). Every step by hand is in the guide; every command, in docs/commands.md.
## How it works
It reads your own records. About 30 sources, each on its own: WhatsApp, Slack, Gmail, Microsoft 365, the Mac's Mail and Calendar, Google Takeout, Apple Health and patient portals, Spotify, Strava, notes, PDFs and other exports. Re-reading one changes nothing. Some readers expect a particular export format (docs/datasources.md).
One model of you, in text you can read. Observers read your side of each conversation, a month at a time, and write notes about you. The master folds them into one self-model of your interests, voice and routines, which you can read and correct. Small agents, "clones", each read it and do one job: tasks, where your care stands, a feed of papers, a map of people, a searchable memory (docs/architecture.md).
It drafts, you send. A reply for each chat waiting on you, written the way you write to that person, and emails, plans, call scripts and filled-in forms. You press send. The two exceptions are yours to start: an outbox you switch on per app, only where its drafts have gone out nearly unchanged many times (docs/memory.md), and a phone call you place, where it says it's automated and never agrees to pay (docs/calls.md).
Your work and your life. For each project or paper it writes where things stand and, given the code, does the next step on a branch for you to accept. It also keeps your job leads, a journal of each day, a diary of your life by month, and your weeks of listening, movement and messages over time.
收录徽章(可选)
想把收录标记放进 README 就复制这行,可选,不影响收录。
[](https://anclum.com/zh/s/mindemory-proxenos/)
- 来源
- github · https://github.com/mindemory/proxenos
- 按领域浏览
- 开源项目
- Traction
- 4 ★
- 收录于
- 今天
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。