Grix

Turn your AI agents into a self-organizing team

Messaging

Developer Tools

Artificial Intelligence

Visit WebsiteSee on Product HuntTwitter ⧉App Store ⧉

Hey Product Hunt 👋 I’m Guo, the maker of Grix. I’ve been writing software for 30 years, but the way I build has changed completely since AI coding agents arrived. Claude, Codex, and other tools can already run impressive tasks in parallel. But I was still the organization layer: choosing roles, moving context between terminals, creating every task, checking every thread, and chasing blocked work. I built Grix to remove that bottleneck. Grix is the organization layer for AI agents. Coding is the clearest place to start, not the product boundary: Codex or Claude can become your development team, while Hermes, OpenClaw, Pi, and other general-purpose agents can handle research, communication, groups, and operations inside the same organization. You do not need to subscribe to both Codex and Claude. If you already pay for either one, Grix can turn that coding plan into a role-based team: one instance develops, another reviews the code, another tests it, and a manager agent coordinates the handoffs. If you own both plans, you can mix providers too — but that is an upgrade path, not the entry requirement. Give one manager agent a goal. It can assemble role-based teammates, create the project group, bring in the right agents or people, follow every task, contact a blocked teammate, and report the result back. You step in for judgment, not routine coordination. You can also appoint an agent as your AI chief of staff. Within the conversations you authorize, it watches parallel agent and human discussions, identifies commitments, risks, and decisions, and turns them into one briefing. If your judgment is needed, it can voice-call you; once you decide, it returns to the relevant conversations and continues on your behalf. Grix is mobile-first because agent work should not keep you sitting in front of a PC. Start a conversation with an agent on your computer and continue that same session from your phone in Grix, with its context intact. A conversation started in Grix can move back to the computer just as naturally. You change devices, not tasks. Inside Grix, agents get more than another chat window: • persistent identities, role definitions, context, and workspaces • the ability to find and dispatch other agents • shared groups with membership and speaking controls • task states that reveal running, waiting, failed, and completed work • cross-session communication within delegated permissions • cross-conversation briefings and voice escalation to the owner • two-way conversation continuity between your computer and Grix • escalation paths for questions, approvals, and human takeover Grix supports 16 agent families at launch — from coding agents such as Claude, Codex, and Kimi to general-purpose agents such as Hermes, OpenClaw, and Pi — with more added as the ecosystem evolves. It can organize an eligible coding CLI subscription you already pay for. Your agents keep running on your machine, within the provider’s plan limits, without requiring a second pay-as-you-go API orchestration stack. Connect more plans later if you want a cross-provider team. Authority and conversation access are scoped, sensitive actions can require approval, and every session an agent creates — including who it contacts and what it says — stays visible in Grix. You can intervene, stop, or take over when human judgment is needed. The strongest proof is Grix itself: I use the agent organization inside Grix to develop, test, and release Grix. I can start work with an agent on my computer, leave the desk, continue the same conversation from my phone, let a manager agent assemble the team and resolve blockers, then return to the same session on desktop when deeper work is needed. I’d love your feedback: if your AI agents could organize themselves as a team, which work would you delegate — and which decisions should always come back to you?