Laya reads your notifications so you don't have to
A local-first desktop app that intercepts Slack, Gmail, Jira, and GitHub events, has LLM agents research them, and hands you ready-to-approve Action Cards.

What it does
Laya aggregates notifications from nine-ish professional tools (Slack, Gmail, GitHub, Bitbucket, Jira, Linear, Notion, Outlook, Google Calendar), runs each event through an LLM pipeline that classifies, researches, and stages a proposed action, and shows you a card you approve or dismiss. Approved actions — replies, PR comments, emails — go back out through n8n. It runs locally via Ollama or LM Studio, or against cloud APIs with your own keys.
The interesting bit
The architecture is a three-layer sandwich: n8n normalizes inbound events, a Python engine does the LLM work, and a Tauri/Svelte desktop shell shows the cards. The clever part is the multi-persona routing — events go to specialized workers (Engineer, Comms, Ops, Sales, HR, Finance) with domain-specific prompts and tools. It also learns from your corrections, extracting classification and context-linking rules over time.
Key highlights
- Local-first: Ollama/LM Studio for the whole pipeline, with BYOK cloud models (Claude, GPT, Gemini) via LiteLLM as an option
- Card Workspaces let you collaborate with a coding agent (Claude Code, Gemini CLI, Codex, etc.) on multi-step tasks like bug fixes
- Hybrid search: ChromaDB vectors fused with SQLite FTS5/BM25 via Reciprocal Rank Fusion
- Budget tracking with monthly cost caps and automatic pause; agent inference backends can even piggyback on a CLI agent’s existing quota
- Coherence traces one entity (a ticket, a person, a PR) across all platforms with AI-generated narratives
Caveats
- It’s a big machine: Python engine, n8n instance, Tauri shell, SQLite, ChromaDB — expect a heavyweight local footprint under
~/.laya/ - The README is feature-dense but quiet on how well the autonomous “research and action-staging” actually performs in practice
- Linux dev builds have a documented inotify quirk that silently breaks Tailwind CSS generation
Verdict
Worth a look if you’re drowning in cross-platform notifications and want an AI triage layer that stays on your machine. Skip it if you want something lightweight — this is a full command center, not a notifier.
Frequently asked
- What is aayushch/laya?
- A local-first desktop app that intercepts Slack, Gmail, Jira, and GitHub events, has LLM agents research them, and hands you ready-to-approve Action Cards.
- Is laya open source?
- Yes — aayushch/laya is open source, released under the Apache-2.0 license.
- What language is laya written in?
- aayushch/laya is primarily written in Python.
- How popular is laya?
- aayushch/laya has 834 stars on GitHub.
- Where can I find laya?
- aayushch/laya is on GitHub at https://github.com/aayushch/laya.