---

What is OpenLoomi?

Your work is scattered across apps. OpenLoomi connects the tools you use and, with your permission, understands what’s on your screen. It brings together the context behind your people, projects, and decisions, then surfaces only what needs your decision or action.

Focus on what matters. This is your Attention Agent.

What is it for?

An always-on desktop attention agent — the friendly desk companion Loomi — watches the door so you don't have to, and turns the day's scattered signals into decision cards you can approve in one tap. Use it standalone, or plug any Agent framework into the same resident desktop: Claude Code, Codex, OpenCode, Hermes, and OpenClaw all work.

• Stop dropping commitments. Overdue replies, creeping deadlines, "I'll send that follow-up Friday" — Loomi nudges you with a small bubble at the right moment, and you can teach it the signals and decisions to watch 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 every tool and channel, no more digging through Slack, Gmail, or Notion.
• Morning brief at 9 AM, end-of-day recap at 6 PM. Today's to-do slides into view every morning, today's done ships every evening — the key info lives in one place instead of nine apps.
• 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.

→ Learn more about the always-on companion in the Attention Agent docs and Use cases.

Features

Quick Start

Download directly (for end users):

Full documentation: openloomi.ai/docs

Use as an Agent plugin (for Claude Code / Codex users):

OpenLoomi ships official marketplace plugins that turn your existing agent into a front-end for the local OpenLoomi runtime.

The slim public marketplace lives at melandlabs/plugins so adding it only fetches the plugin payloads. See the plugin docs for full reference: plugins/claude · plugins/codex.

Use in skill-only agent runtimes:

Install the OpenLoomi skill set directly from this repository:

npx skills add https://github.com/melandlabs/openloomi/tree/main/skills \
  --skill openloomi openloomi-setup openloomi-memory openloomi-connectors openloomi-loop openloomi-goals openloomi-api openloomi-feature-guide composio \
  -y

Start with the openloomi or openloomi-setup skill. The first setup pass checks whether OpenLoomi Desktop is installed, whether the local API is reachable, and whether a local session token is available. If OpenLoomi Desktop is missing, the skill points the user to the official Getting Started guide and release downloads before continuing.

Develop locally (for developers):

git clone https://github.com/melandlabs/openloomi.git
cd openloomi

pnpm install
pnpm tauri:dev

Requires Node.js 22+, pnpm 9+, Rust 1.75+, and on Windows: Visual Studio Build Tools with C++ workload. For more platform-specific setup requirements, see CONTRIBUTING.md.

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.

App Screenshots

Feedback

This is early-stage software. We're looking for people who'll actually install it, connect their tools, and tell us what's broken.

• GitHub Issues — bugs, install problems, feature requests
• Discord — discussion, questions, help
• Email — anything else

Contributing

See CONTRIBUTING.md. Look for good first issue labels.

Related Projects

OpenLoomi is the desktop experience. Many of the lower-level building blocks it depends on — AI runtime, integrations, loop engine, audit, memory — are developed in the open at:

OpenContext — the context layer

Your agent forgets why it made decisions. OpenContext fixes that.

OpenContext is the agentic context runtime that sits underneath an agentic application — a temporal context graph, a memory API, retrieval primitives, a multi-platform integration mesh, a deterministic loop engine, and an agent runtime, all behind a single library-first dependency.

• Temporal Context Graph. A directed acyclic graph where every fact carries valid_from / valid_until. Supersession, contradiction, and merge are first-class edges — corrections are append-only, not destructive, so "what did we believe last quarter?" is a real, citable query.
• Library-First API. Drop it into any host process with pnpm add @melandlabs/opencontext. It's not a UI, a chat surface, or a model provider — it's the glue between durable memory, retrieval, context correction, and multi-platform connectivity.
• Platform Integration Mesh. One uniform IntegrationRecord shape across Gmail, Slack, Telegram, Linear, Jira, iMessage, Feishu, Weixin, and friends — credential rotation, rate-limit handling, and reconnect logic live behind the adapter.
• Deterministic Loop Engine. A scheduler that wakes up, decides whether there is real work, and only then calls into the agent runtime. LLM calls are not the foundation — they are the last step.
• Built for embedded agents. Designed to be embedded into any host process — CLI, MCP server, daemon, desktop app, or product backend. OpenLoomi Desktop is one such host.

→ Read docs/architecture.md for the full data model, lifecycle of a fact, and transport surface map.

pnpm add @melandlabs/opencontext

License

Apache 2.0