Don't deploy an agent from scratch. Collaborate with a professional team in one click.

English | 中文

_{Found a bug or need help? · Open an issue · Discussions}

Agents that already know how to remember, collaborate, and use tools — start from a template, or compose your own.

A 90-second tour — install, core concepts, and templates in action.

See more templates →

NarraNexus is a multi-agent product — not yet another framework where you wire agents together, but a ready-to-run team of agents that already collaborate. Three deployment paths — pick whichever suits you.

1. Open agent.narra.nexus
2. Sign up
3. Pick a template and go

The app bundles its own runtime — no Python / Node / Docker to install.

1. Download the app
2. Drag it into the Applications folder
3. Launch → pick a template and go

git clone https://github.com/NetMindAI-Open/NarraNexus.git
cd NarraNexus
bash run.sh

run.sh checks prerequisites (uv / node / tmux) and launches all local services. For the full service/port list and detailed setup, see the dev docs.

Three doors, same place.

Reference implementations — use them directly, or fork to customize.

For investors and analysts who read markets at 7am. 6 agents deliver an analyst-grade HTML briefing to your inbox daily at 08:00 Asia/Shanghai. Not yet another news summary — a structured read on "what is the market trading today, and should I attack, defend, or watch?"

narra.nexus/templates/financial-morning-briefing →

For content creators juggling inbound sponsorships. 4 agents parse incoming sponsor emails, keep your CRM up to date, and monitor brand mentions across social platforms — so you spend time on the next video, not on inbox triage.

narra.nexus/templates/kol-assistant →

For teams juggling internal team chat and external client communication. A single bot maintains two searchable knowledge bases — internal-only and client-shared — and auto-files every chat, doc, and meeting note into the right scope. Tone adapts per audience; language is auto-detected.

narra.nexus/templates/pm-bridge-bot →

All built by NarraNexus agents themselves.

• LLM API key: the cloud version has a free trial quota. For daily or local use, you'll want your own LLM API key — one or two minutes to register at NetMind, OpenAI, or Anthropic.
• An agent isn't 100% out of the box: it needs your corrections and feedback to get sharp. Treat it like a new hire, not a god.
• Collaboration isn't perfect on the first try: complex tasks may take a couple of iterations before agents get the hang of it; some judgment calls always belong to you — the team handles the bulk, the key calls stay with you.
• Architecture trade-off: each agent launches its own MCP process, adding ~100ms of startup overhead. Negligible for chat-style workflows; switch to Direct Trigger mode for high-frequency jobs.

• Discord — discord.gg/ReCMd6a2wf
• Twitter / X — @NetMindAI
• Email updates — coming soon
• Feedback — GitHub Issues

NarraNexus is built to work well with human and AI-agent contributors alike.

Start here:

• New contributors → CONTRIBUTING.md
• AI coding assistants → AGENTS.md (vendor-neutral) or CLAUDE.md directly
• Project map for your AI agent → .mindflow/_overview.md

How the project is run:

• Governance & maintainer team → GOVERNANCE.md, MAINTAINERS.md
• Community standards → CODE_OF_CONDUCT.md
• Security policy → SECURITY.md

See MAINTAINERS.md for the current maintainer team. Run git shortlog -sn for the full contributor list.

CC BY-NC 4.0