What is OpenLoomi?
Your agent can read the update. OpenLoomi helps it understand what changed, what still matters, and what should happen next.
Open-source, local-first, model-agnostic context runtime. Plugs into Claude Code, Codex, OpenCode, Hermes, and OpenClaw — no need to replace your model or your agent.
The Attention Agent that runs across every Agent Runtime
OpenLoomi is the open-source, local-first Attention Agent for developers and small teams building long-lived working agents on Claude Code, Codex, OpenCode, Hermes, and OpenClaw.
It lets your existing agent continuously understand the people, projects, events, relationships, and state changes that span sessions and tools — and writes each step's outcome back into the next context slice.
It turns the day's scattered signals into real decisions, and keeps orchestrating execution until you have a result.
At a glance
| If you want to… | OpenLoomi gives you… |
|---|---|
| Stop dropping commitments (overdue replies, creeping deadlines) | The Attention Agent — small bubbles on your desktop that nudge without fighting for focus |
| Find anything from your work life in seconds | Memory — a holistic knowledge graph across tools and channels |
| Replace "open nine apps every morning" | Loop — morning brief + decision cards in one place |
| Chat with AI inside Telegram, WhatsApp, iMessage, QQ, or Lark/Feishu | Messaging Apps |
| Schedule recurring work — daily digests, weekly reports, reminders | Automation |
| Plug into an existing Agent shell (Claude Code, Codex CLI, OpenCode, Hermes, OpenClaw) | Plugins and Agent Runtimes |
What is it for?
OpenLoomi gives you a desktop assistant for any Agent framework — Claude Code, Codex, OpenCode, Hermes, and OpenClaw all plug into the same resident desktop.
- Stop dropping commitments. Overdue replies, creeping deadlines, "I'll send that follow-up Friday" — and, more importantly, the follow-up three days later that quietly supersedes it. The desktop bubble nudges you, but the system underneath is doing more than recall: it maintains the current state of every commitment, with the original message and every transition still auditable. You can extend the signals and decisions it watches for.
- Find anything from your work life in seconds. "What did I decide about pricing last quarter?", "Who owns design on Acorn?", "What was I working on before vacation?" — memory spans tools and channels, no more digging through Slack, Gmail, or Notion.
- Get a morning brief and an end-of-day recap instead of opening nine apps. The key info lands in one place.
- Ask AI for work help right inside your chat. Draft a reply, summarize a long thread, schedule a follow-up — directly in Telegram, WhatsApp, iMessage, QQ, or Lark/Feishu.
See the Attention Agent doc for the always-on companion that surfaces those reminders, or browse Use cases for end-to-end workflows.
Core Capabilities
| Capability | What it does | |
|---|---|---|
| 🐾 | Attention Agent | An always-on desktop companion (Loomi) that surfaces pre-decided reminders — 9 AM to-do, 6 PM recap, overdue replies — as small bubbles, without fighting for your focus. |
| 🧠 | Holistic Context | Short → mid → long-term memory that grows on its own — visible, auditable, and always remembering your people, projects, and decisions across months. |
| 🔌 | Platform Connectors | Auto-fetch background sync pulls commits, issues, emails, and docs proactively into your context graph. Messaging apps — Telegram, WhatsApp, iMessage, QQ, Lark/Feishu — let you chat with AI directly inside your existing conversations. |
| ⏰ | Proactive Tasks | Schedule recurring work — daily digests, weekly reports, reminders — that run automatically on your desktop. |
| 🖥️ | Security & Local-First | Native app for Windows, macOS, Linux — works out of the box, minutes to set up; local-first storage, AES-256 encryption, no data leaves your machine, auditable access logs. |
| 🧩 | Any Agent Integration | OpenLoomi's context, memory, connectors, attention agent, and Loop engine are all delivered as open-source Skills and Plugins. Use OpenLoomi Desktop directly, or plug into your existing Agent — Claude, Codex, OpenCode, Hermes, or OpenClaw. |
Why It Is Different
OpenLoomi is open-source and neutral. Instead of locking you into a single vendor's agent, it integrates with any Agent Runtime — Claude Code, Codex, OpenCode, Hermes, OpenClaw — and brings a shared, cross-agent layer to all of them: a resident desktop attention agent, holistic context memory, platform connectors, and proactive tasks. Whichever agent you run, OpenLoomi watches the door, remembers what matters, and surfaces only the decisions worth your time — so you spend your attention on the work, not on chasing it.
Explicit, on the record — so the framing is clear before you read the rest of the page:
- Not an LLM. OpenLoomi doesn't ship a model. It routes prompts to your existing Claude / OpenAI / local runtime. See Getting Started → AI Configuration.
- Not a replacement for Claude, Codex, or any other agent. OpenLoomi is a context runtime that sits behind them as a plugin or skill. Your agent stays in charge.
- Not another agent framework. It doesn't compete with Claude Code, Codex CLI, OpenCode, Hermes, or OpenClaw — it complements them. See Agent Runtimes.
- Not a RAG library. Retrieval-augmented generation finds similar text. OpenLoomi maintains the current state of your work and surfaces transitions, not just similar documents. See Memory → How this differs from RAG, vector memory, and chat history.
- Not a knowledge graph in the classical sense. There is no upfront schema, no ontology to maintain. Connections emerge from use (Hebbian-style) and decay when unused. See Memory → Spatial Dimension: Hebbian Memory Connections.
- Not chat-history memory. A flat log of past turns can't tell you that "pushing to next week" superseded "I'll have it Friday" — and a tracker that holds both as facts is worse than no tracker. OpenLoomi models commitments as a state machine with supersession.
- Not a generic automation or notification tool. It does not "remind you of things." It surfaces decision cards backed by attributable evidence and waits for your Approve before any external action.
| Compared with... | OpenLoomi adds |
|---|---|
| Claude Code / Codex / OpenClaw / Hermes | a state-aware context runtime that plugs into your existing agent shell — no model swap, no framework lock-in. See Memory → Commitment Lifecycle for what "state-aware" means. |
| RAG / knowledge bases | work state, not just document retrieval: what changed, what is still true, and what should affect the next action |
| Chat-history memory | attributable, time-bounded, supersedable state with source evidence — not just a flat log of past turns. See Connectors → Trust Ladder for the access model. |
| Claude Cowork-style agents | the open-source, local-first counterpart that maintains the state of commitments (not just their memory), with explicit supersession and Approve-before-action. Learn more → |
Deep Dive
🐾 Attention Agent — Your Always-On Desk Companion
It's 9 AM — you've just sat down at your desk, your inbox isn't even open yet, and the little fox (Loomi) in the bottom-right of your screen pops up a bubble: "Sarah's meeting invite still needs an RSVP — due by 11 AM today." That's the Attention Agent. It doesn't steal your focus; it just gently nudges you in the moments that matter.
-
🪟 An always-on-top, transparent little fox — 168×168, doesn't disappear when you close the main window, hides in the system tray and comes back with a single click.
-
🔔 Only surfaces when it should — an email that needs a reply, a meeting past its RSVP deadline, 9 AM rolling around — a single gentle nudge, nothing more.
-
✋ You can interact with it directly — click to open details, drag it anywhere on screen, right-click to pause, double-click to summon the main window.
-
🎭 Sleep mask, wagging tail, little hops — rich expression animations make it feel "alive" rather than a cold notification.
🔄 Loop — Scans Your To-Dos Every 10 Minutes
At the start of the day, what you really want to know isn't "how many unread emails do I have", but "what do I actually have to personally handle today". That's exactly what Loop is for. Every 10 minutes it picks out the signals scattered across your email, calendar, PRs, and @mentions, prioritizes them, and turns them into "decision cards" placed in front of you — you glance at the preview, agree, tap "Approve", and Loop acts on your behalf.
-
📥 See — pulls signals from email, calendar, GitHub, Slack, and Linear that are relevant to you
-
🧠 Understand — uses OpenLoomi memory to add context to each signal (who the sender is, which project, what you last discussed)
-
🪪 Decide — tags each signal with a type (
email_reply/rsvp/review_pr/im_reply/todo) and a confidence score -
🛡️ Protect — Dry-run always comes first, execution always waits for your Approve
🧠 Holistic Context
OpenLoomi builds persistent knowledge graphs of people, projects, and decisions. Unlike other AIs that forget after each conversation, OpenLoomi's memory grows on its own — completely visible and auditable. Six months later, it still remembers your commitments and context. No more repeating yourself, no more context loss.
-
Remembers Q3 partnership direction confirmed with Sarah 3 months ago (budget: $50K)
-
Remembers demo feedback 6 weeks ago (needed security compliance materials)
-
Today when Sarah emails, automatically pulls historical context and generates reply draft
-
You can always see and audit exactly what OpenLoomi remembers about you
💬 Chat — Assign Tasks Like Talking
Describe your needs in plain language; OpenLoomi automatically understands intent and breaks it into executable steps. No complex commands to learn. Learn more →
"Help me prepare investor meeting materials for next Monday"
OpenLoomi: "Got it. I've compiled the past 3 months of funding progress, created a competitive analysis, and generated a 10-page PPT framework. Want me to send it for your review?"
📱 Messaging Apps — OpenLoomi in Your Favorite Apps
Use OpenLoomi directly inside Telegram, WhatsApp, iMessage, QQ, and Lark/Feishu — no separate app required. Assign tasks, get alerts, and receive results right where your conversations already happen. Learn more →
🔒 Security & Local-First — Your Data, Your Sovereignty
Privacy and intelligence aren't mutually exclusive. OpenLoomi uses AES-256 end-to-end encryption, hardware-isolated processing environments without any public gateways, and zero training commitments — your data never trains public AI models. Local-first architecture keeps raw data on your device, while full audit logs and SOC 2 compliance (in progress) give you complete visibility into every data access. Learn more →
🧩 Any Agent Integration — Skills + Plugins
OpenLoomi's context, memory, connectors, attention agent, and Loop engine are all delivered as open-source Skills and Plugins. Use OpenLoomi Desktop directly, or plug into an Agent you already use every day.
Skills catalog covers the everyday work scenarios — code generation, PDF creation, data analysis, browser automation, document creation, email writing, image generation, PPT creation, deep research — and is continuously growing. Import any skill from a GitHub repo or the Skill Hub.
Plugins turn OpenLoomi's runtime into a front-end for existing agents: Claude Code, Codex CLI, OpenCode, Hermes, OpenClaw. Install once, and your agent gains OpenLoomi's holistic context, attention agent, and loop engine.
Inside Claude Code — once the OpenLoomi plugin is installed, type /openloomi: and the slash-command autocomplete surfaces every OpenLoomi command (/openloomi:setup, /openloomi:status, /openloomi:memory, /openloomi:loop, …):
Inside Codex CLI — once the OpenLoomi plugin is installed, type @OpenLoomi (or just @O) and Codex surfaces the OpenLoomi skill namespace (@OpenLoomi setup, @OpenLoomi memory, @OpenLoomi loop, @OpenLoomi connectors, …):
- Browse the Skill catalog →
- Browse the Plugin catalog →
- See how OpenLoomi's surface fits into existing Agents →
A Complete Intelligence Loop: From Perception to Action
OpenLoomi implements a full "Receive → Process → Remember → Understand → Serve" loop — mirroring how a capable human colleague operates:
| Layer | What It Does |
|---|---|
| 📡 Receive | Multi-platform, multi-modal ingestion — IM, email, documents, files, web data, voice calls — unified into one pipeline |
| ⚙️ Process | Noise reduction at scale — deduplication, OCR, ASR, intent extraction, semantic clustering — turning chaos into structured signals |
| 🧠 Remember | Persistent knowledge graph — people, projects, decisions — survives conversations, platforms, and months of time |
| 🎯 Understand | Deep semantic intent, not just keywords — cross-modal understanding, emotional tone, contextual relevance |
| 🚀 Serve | Proactive delivery — smart summaries, auto-replies, scheduled reports, personalized alerts — without being asked |
Use Cases
🌍 Global Managers
Never miss a critical signal across time zones. Business runs 24/7 globally. OpenLoomi filters time zone and language noise, capturing high-value opportunities while you sleep — wake up to a refined action list. Read the full story →
🧑💻 Engineers & Product Teams
Team memory that never decays. Transform discussions scattered across Slack, Jira, and documents into structured knowledge. Auto-generate weekly reports, sync missed context, eliminate "context rot." Read the full story →
🚀 Founders & Salespeople
One person does the work of many, at scale. OpenLoomi learns your communication style, automatically maintains hundreds of client relationships, follows up on leads, generates personalized proposals — never burns out. Read the full story → · Salespeople →
🛠️ Automated Dev Reports
Stop wasting hours on manual dev reports and issue triaging. Every night at 10 PM, OpenLoomi automatically pulls commits, compiles a structured dev report, emails the team, fetches new GitHub issues, rewrites them with full context, and syncs everything to Linear. What used to take hours per week now runs itself. Read the full story →
📱 Social Media Managers
Run your X (Twitter) account on complete autopilot. Every 4 hours, OpenLoomi researches trending topics, engages with relevant accounts, posts approved content, and tracks every action — while you simply approve before anything goes live. Build an engaged following 24/7 with just ~10 minutes of your time per day. Read the full story →
See All Use Cases →
Related
- Getting Started — install + first run
- Attention Agent — the always-on companion
- Loop — proactive decision cards
- Memory — holistic context and semantic search
- Plugins — bridge into Claude Code or Codex CLI
- Privacy & Security — how OpenLoomi protects your data
- Use Cases — real workflows in context