Meet EverOS Cloud. 💛
Your agent shouldn’t have to start from scratch every time a conversation ends.
The preferences a user shared. The decisions made during a project. The approach that worked on a previous task. That context should be available when it matters again.
EverOS Cloud gives your agents persistent memory through an API:
• Remember users and retrieve relevant context across sessions
• Turn task experience into Cases and reusable Skill memories
• Bring conversations, images, and documents into your memory workflow
You build the agent experience. We handle the memory infrastructure, so you can spend less time maintaining the stack and more time making your product useful.
Subscribe to EverOS Cloud Pro and get 3× the standard Pro usage allowance for 6 months.
Start here:
everos.evermind.ai
Some guide. Some build. Some create. Some spread the word.
Together, you keep our open-source projects moving forward.
To our advisors, builders, creators, and evangelists.
Thank you for being part of the EverMind Open-Source
Ambassador community.
Your time, ideas, and contributions mean a lot to us. 💛
@li9292 @Tz_2022 @xiqingongzi @hylarucoder @OipsAnthony @yinmin1987 @t20000622yy @Av1dlive
EverOS open-source v1.4.2 is here.
This release tackles problems that show up when memory servers keep running:
• Fixed a file-descriptor leak in long-lived indexes
• Separated query embeddings from background writes, so searches don’t queue behind ingestion
• Fixed cross-project index overwrites, your Markdown memories were never affected
Upgrade: pip install --upgrade everos
Check the release notes before upgrading: Milvus and certain multi-project deployments require an index rebuild.
Prefer managed memory infrastructure?
Build with EverOS Cloud.
evermind.ai/everos
Your agent finished the task. What should it keep for next time?
Not just the conversation, the approach that worked.
EverOS captures task execution as Cases and consolidates recurring successful approaches into reusable Skill memories. Agents can retrieve those procedures when a similar task comes up.
Memory for what happened. Skills for how to do it again.
Build with EverOS Cloud. We handle the memory infrastructure, you focus on your agent.
everos.evermind.ai
Trying a new personal agent is exciting. Uploading the same documents and explaining your work all over again is less exciting.
Over time, your agent picks up the small things: how you like things written, what your project is about, which decisions you’ve already made. Moving that context elsewhere takes work, and some of it gets lost along the way.
With EverOS, we’re giving that context a place to grow across supported agents. Save your conversations, documents, and useful knowledge, then connect your agents to the same memory source.
We recently shared how to connect EverOS to Muse. You can start building your own memory today and keep adding to it as your work evolves.
Sign up for free: everos.evermind.ai
In case you missed it: EverOS open-source v1.4.1
The 1.4 series brings native Windows support, faster memory retrieval, and more reliable indexing. The v1.4.1 patch fixes a dependency issue affecting fresh installs.
Building an agent that needs to remember users, retrieve past context, and reuse lessons from previous tasks?
EverOS Cloud provides managed memory through an API, with support for conversations, images, and documents. We handle the memory infrastructure so you can focus on your product.
Start building for free:
everos.evermind.ai/
EverOS, meet @Muse 💛
Connect EverOS to Muse as a custom connector to save and retrieve context from your projects, past decisions, useful knowledge, and task outcomes.
A personal agent, with another memory source to work from.
Setup instructions, give this link to Muse:
evermind.ai/connectors/muse.…
Love seeing our open-source ambassadors help more developers get started with EverOS and long-term agent memory.
Thanks for putting this walkthrough together, @yinmin1987 💛
Maybe Raven 🐦⬛ can lend a wing with the next video.
First Daily. Then Weekly. Now Monthly. 💛
Raven 🐦⬛ reached #1 on @HuggingFace @HuggingPapers Daily and Weekly Papers, and now tops September’s Monthly Papers list.
We’re building the “harness of harnesses”: coordinating specialist agents, preserving experience through EverOS memory, and turning that experience into reusable skills.
Open-source code you can run, research you can examine, and a community helping us move forward.
Thank you to everyone who read the paper, tried Raven, and supported the project. This means a lot to our team.
github.com/EverMind-AI/raven
Thank you for all the love for Raven 🐦⬛ v0.2.0, the harness of harnesses, built for RSI. 💛
We kept shipping through China’s National Day holiday, and v0.2.4 is already here.
You can now ask Raven to inspect and update its own settings in chat, with approval controls and API keys kept out of the conversation.
Plus: unified Connectors, clearer approvals, and workspace diffs.
Huge thanks to the team and everyone building with Raven.
github.com/EverMind-AI/Raven
EverMind retweeted
And we are live with another rendition of @MiniMax_AI intelligence in the open with @FireworksAI_HQ, @digitalocean, @radixark, @Modular and @evermind
EverMind retweeted
拥有 Harness of harnesses,以及 RSI (递归自我进化)架构的 Raven,国庆节也要休息休息啦!
欢迎大家持续关注 @evermind 最新 harness 架构的更多动态哦!目前我们的技术报告也已经在 arxiv 上发布啦!欢迎大家查看!
模型一直在升级,是不是可以不需要给 「Agents」当项目经理了?
前几天我给大家分享了 Raven 做的游戏、PPT,还有调研和实验案例,在不同程度上说明了现在多 Agent 的能力边界,无论是视觉效果还是功能,都很强大。
现在让 AI 完成一项具体工作,已经不稀奇了。但真正做项目时,我们的精力却经常花在安排顺序、同步信息和检查结果上。
研究报告生成了,要挑出有用的结论交给写代码的 Agent;代码改了,要通知负责测试的 Agent;设计调整了,还要确认其他成员有没有用错版本。
每个工具都节省了一些时间,管理这些工具又成了一份新工作。
这就是 Raven 希望解决的问题:当一项任务需要多个专业 Agent 参与时,怎样减少协调成本,让整个项目持续推进?
多 Agent,其实有不同的组织方式。如 图1 所示。
一种是任务之间相对独立。比如调研几个不同的项目,可以同时执行,最后汇总结果。这类独立并行的协作方式,可以称为 Swarm。
另一种存在明确的前后依赖。比如代码实现完成后才能进行相应的集成测试,游戏资产准备好后才能完成集成。这类按依赖关系组织的方式,通常采用 Graph(图式编排),其中常见的一种结构就是 DAG(有向无环图)。
还有一种强调长期协作。不同成员持续负责研究、开发、审查等工作,保留各自的上下文和经验,在项目中反复参与。这类组织方式可以称为 Agent Team(智能体团队)。
这些方式可以组合使用。选择哪一种,取决于任务有没有依赖、上下文是否需要隔离,以及成员是否需要长期保留经验。
Raven 的任务编排,首先处理其中的依赖问题。
DAG 是工作流系统中常见的结构。每个节点代表一项任务,箭头表示任务之间的依赖。
拿我们之前展示的游戏来说,其中一条资产制作流程包含八个节点:
参考研究、现有资产审查、美术规范、武器制作、Boss 制作、资产集成、构建验证,以及游戏内效果审阅。
参考研究与资产审查可以同时进行。美术规范确定后,武器与 Boss 可以并行制作。集成则必须等到资产准备好。
无环不代表不能修改。一个开发节点内部可以多次测试和修复;当前一轮完成后,也可以根据结果安排新的任务。无环约束的是图中的依赖关系。
节点也不必与 Agent 一一对应。同一个成员可以承担多个任务,没必要为每一步都创建一个新角色。
执行顺序确定以后,还有一个容易被低估的问题:交接。
研究节点显示已完成,开发节点也并不能开始工作。它需要知道采用哪种方法、结论有什么依据、哪些条件还不确定,以及具体该读取哪份文件。
缺少这些信息,后续成员可能重复研究,也可能把一个有前提的结论当成通用事实。
因此,Raven 在调度任务时,也会组织明确引用的上游输出和产物入口。前置任务失败时,依赖它的节点可以被跳过,不相关的独立分支继续执行。
这也涉及上下文管理。研究成员不需要读取所有调试日志,设计成员也不需要了解每一次参数实验。每个成员获得与当前任务相关的信息,同时保留查阅上游依据的入口。
我们希望减少的不只是等待时间,还有重复阅读、重复判断和信息误传。
当然,状态上的“完成”不代表质量已经合格。验收标准仍然需要提前定义。
Harness of Harnesses,是在这个基础上的进一步选择。
Harness 是支撑 Agent 执行任务的运行机制,包括上下文管理、工具调用、规划、检查和错误处理。
不同专业 Agent 已经形成了不同的工作方式。Research 需要检索和证据组织,Code 需要开发与调试,Design 需要视觉制作与交付,Oncall 需要实验运行和持续观察。
Claude Code、Codex 等工具,也各自拥有成熟的执行机制。Raven 希望保留这些专业能力,在更上一层安排成员、任务、依赖和结果交接。成员内部如何执行,则由各自的 Harness 负责。
这就是我们所说的 Harness of Harnesses。
它的价值需要通过完整任务来检验。多一个成员,就会增加一次协调和交接。简单任务交给一个 Agent 完成,往往更合适;复杂任务才需要认真考虑专业分工带来的收益。
但完成一次项目,还不够。
我们更在意的是:这个项目里纠正过的问题,下次能不能少出现一些?
用 AI 久了,很容易遇到这种情况:它能复述你的要求,执行时却还是用了旧方法。
保存一段聊天记录,并不一定会改变后续行为。经验需要进入实际的规划、工具使用和检查过程。
在 Raven 中,EverOS 承接跨会话的上下文与经验,Playbook 保存成员、任务和依赖安排。我们进一步探索的是,让执行中的反馈能够修改 Harness 本身。
这与 RSI,也就是递归自我改进有关:系统利用执行结果和反馈,改进产生这些结果的机制。
这种改进不一定发生在模型参数上,也可以发生在提示、策略代码、工具规则和工作流程中。
例如,一个旅行规划 Agent 经常安排跨区折返。用户指出问题后,后续规划可以增加区域优先的安排方式,并在交付前检查交通与步行负担。反馈如果落实成这样的执行变化,就不需要每次重新提醒。
Raven V0.2.0 中的 Curator 是这方面的实验性实现。它结合任务要求、执行记录和用户反馈生成 Harness 改动,经过声明检查、实际装配与预检后安装;安装失败时,触发恢复流程。
用户提出的要求,由此有机会成为后续任务实际采用的方法。
自我改进最需要验证的,是“改进”这两个字。
修改了提示或代码,只能说明系统发生了变化。当前任务表现更好,其他任务有没有退步?提升是否能重复出现?额外消耗的时间和 Token 是否值得?评价标准能否保持独立?
这些问题不能靠一次漂亮的演示回答。
目前,Curator 已在模拟案例中完成多轮反馈与修改,但跨场景、重复实验和长期效果仍需要继续验证。专业能力评测也不能直接作为自我改进有效的证据。
对 Raven 来说,DAG 负责组织当前任务,专业 Harness 负责执行,记忆和 Playbook 保留可以复用的信息与安排,Curator 则探索如何根据反馈调整工作机制。
这些设计最终应该减少的是人的重复劳动,而不是人的判断权。
目标是什么、哪些行为不能接受、资源有限时如何取舍、什么结果可以交付,仍然需要由人决定。
我们最希望的是一个项目结束时,留下的不只有作品,还有完成这类工作的经验。
下一次面对类似问题,我们能从这些经验继续,而不是重新解释背景、重复同样的纠正。
模型会更新,工具会更换。我们在真实工作中积累的判断,应该能够保存下来,并继续发挥作用。
Raven 0.2.0,带着最先进的理念以及强大的功能,开放给大家!欢迎体验! @evermind
Raven 🐦⬛ reached #1 on Hugging Face Papers’ weekly list.
We’ve spent a lot of time working on how specialist agents coordinate and carry useful experience into their next task. Seeing people pick up the paper means a lot to the team.
Thanks to everyone who read it, upvoted, or tried the code.
Raven is open source. Give it a run. We’d love to hear what worked, what broke, and what you’d build with it.
github.com/EverMind-AI/Raven
Keep using Claude Code and Codex. Bring them into a team with Raven 🐦⬛.
Raven 0.2.0 coordinates them alongside its Research, Code, Design and Oncall harnesses, with a shared task graph and memory across sub-agents.
We’re also experimenting with improving the harness itself: the prompts, policies and strategy code that shape how agents work. That includes the orchestration layer deciding how the team works together.
We’re opening up early collaborations with teams building domain-specific agents to explore and evaluate Raven’s RSI capabilities together. 💛
Raven 0.2.0 — The Harness of Harnesses, built for RSI. 🐦⬛
One harness can't be best at everything. Raven combines its own specialist harnesses (Research, Code, Design, Oncall) with the agents you already use (Claude Code, Codex and more) into one team.
And it's built for RSI, and not just at the skill level. The whole harness can be rewritten by AI: prompts, policies, strategy code, playbooks. Every sub-harness, including the orchestration layer itself, is its own instance that can be improved.
With Raven you can:
1. Orchestrate many agents as one team. Raven's sub-harnesses and external agents work in one task graph with shared memory across sub-agents, powered by leading orchestration (0.963 Node F1 on the Multi-Agent Orchestration Benchmark).
2. Run long, complex tasks. Oncall and proactive execution keep work going for days, from scientific research loops to shipping a full Godot game.
3. Build vertical agents with RSI. Use Raven's RSI to develop and refine an agent for your domain, and we'll optimize it with you. Experimental for now; reach out to the Raven team(Discord:discord.gg/VVSm8MESch).
More in the video and slides below. Open source, Apache-2.0.
github.com/EverMind-AI/Raven (lots of work made with Raven lives there, and much of this launch's material was made with Raven too)
EverOS is now a @convex-dev component. 🧠
Add persistent long-term memory to any Convex app, or to @convex-dev/agent in one line, backed by EverOS Cloud.
Convex already gives agents the backend primitives they need: a reactive database, durable functions, scheduling, and threads. Now agents can remember across threads and sessions too.
EverOS turns conversations into durable user memory: facts with timestamps and traceable sources, plus a profile that evolves over time. That memory belongs to the user and follows them across threads, sessions, models, and agents.
Under the hood, EverOS delivers state-of-the-art results across long-term memory benchmarks.
No vector database to run. No memory infrastructure to stitch together.
Get it today:
convex.dev/components/everos…
















