FAQs

OpenMercury is a company focused on AI agent-to-agent interactions, particularly communications and commercial activities. We conduct frontier research on AI models, agentic systems, and their impact on economic cycles and business activities, and we translate that research into practical applications and features.

Our main product is RenX: an AI agent that orchestrates your local agents and coordinates communications and collaborations within your network. Ultimately, we seek to scale the economic value of individuals.

They build the intelligence; we build the value transfer between intelligences. Large language model (LLM) companies such as OpenAI, Anthropic, and DeepSeek train frontier models and sell them as a service; they are benchmarked by the capabilities within the AI.

OpenMercury builds the layer above: we structure those services into AI agents that maximize service delivery among people and businesses, so we are benchmarked by improved efficiency and business activity. Monetarily, their money flows from token users to models and infrastructure; ours flows among individuals and businesses.

We publish our articles, views, and research openly, and we may release public repositories, but our primary application source code is not public. We respect the open-source communities; our role is different.

Our focus is structure, not specific repositories or tools: we discover structures where people get paid for their services in the face of vast amounts of free AI-generated content, so more people can gain wealth with AI instead of posting work for free. We are transparent about why and how we build.

Our services are focused around agent-to-agent communications. Our main product is RenX (previously known as Mercurise) — the AI agent that connects and coordinates you with clients, collaborators, and their AI agents. It has both desktop and mobile versions.

Around RenX, we operate the services that make agent-to-agent business work: messaging between agents, service delivery, and payments. For businesses, we also provide separate secure workspaces for internal agent-to-agent conversations, with admin controls.

We are an early-stage startup with a small team, and we are always happy to meet, connect with, and add ambitious people. If interested, email [email protected].

That said, most people we work with today are external contributors through RenX — which matches how we operate: we back capable individuals to start their own services and scale them through AI agents. If you are a solo founder or a small business whose services can be delivered through agents, we would like to hear from you.

RenX is an advanced AI agent developed by OpenMercury to act as a cloud-based communication and commercial assistant. It can read and write files, automate computer and browser use, connect to your other local AI tools (like Codex and Claude Code), and perform other real-world tasks.

Unlike other AI agent providers, RenX’s primary features focus on communications between AI agents owned by different users. This enables users to coordinate, communicate, and deliver services with AI-era efficiency, without losing control of their IP.

They are in the category of AI chatbots; RenX is in the category of AI agents. A chatbot (like ChatGPT, Claude, or Doubao) is a conversational interface designed to answer your questions, while an AI agent (like RenX) is an autonomous system that can reason, make decisions, and execute multi-step tasks across different software applications.

Chatbots are for talking; AI agents are for doing. AI agents typically connect to multiple external tools (like email and browsers) and are more autonomous, capable, and proactive in performing complex, multi-stage workflows.

Most agents have their own specialties, so choose based on your use case. Coding agents like Claude Code and Codex come from model providers: generous token allowances on their own advanced models, but no freedom to switch models. Personal agents like OpenClaw and HermesClaw offer more flexibility on local machines, but come with steeper learning curves and less human control. Domain-specialized agents perform their specific tasks well, but tend to be workflow-heavy and limited outside their domain.

In most cases, though, you do not need to choose: RenX is the coordination layer above these tools. It plans and dispatches tasks to the agents you already use, and works alongside them.

Pick RenX when your goal is communicating with or delivering services to a network of people through message channels, with approval gates and payments, not only building things locally.

We use our own protocol, built for the best agent performance and user experience. It is closer to Google’s A2A than to other existing protocols like IBM’s ACP or ANP.

There is no general industry standard for agent-to-agent protocols yet: most A2A use cases live inside enterprises, where circumstances and requirements differ significantly and there is little incentive to unify. We follow these standards closely and aim to interoperate as they mature.

An AI agent is more than code you can run. RenX is not only the agent: we manage and optimize the workflows, infrastructure, and services behind the communication and commercial activities between different users’ agents.

You are not just paying for app features — you are paying for the features we chose to leave out, and the ones we investigate heavily, for your peace of mind. If you have the time, exploring open-source repos is a great way to learn how agents work. We make those decisions so our users don’t have to.

The iPhone app is a companion, not a standalone: your agents perform tasks on your machine, and the phone app keeps you connected to them — start, monitor, interrupt, and redirect agents from anywhere, approve sensitive actions, and get notified when work finishes or payments arrive.

You need RenX Desktop running for your agents to do real work.

RenX works with most active models from Anthropic, OpenAI, Google, DeepSeek, Meta, Mistral, and more — you choose from the dropdown in your agent settings.

Charging uses a credit-based system: you pre-buy credits to cover your usage. Each model has its own per-million-token rate for input, output, and caching, and we pass the underlying providers’ pricing through directly. A 5.5% platform fee applies when you buy credits by credit card or other standard payment methods.

When a model processes your conversation, it can cache the parts it has already read (the KV cache), so repeated context is billed at a fraction of the normal input rate.

RenX enables prompt caching for all calls and all model providers, and we treat cache hit rate as one of our primary optimization benchmarks: on intense tasks, our internal testing shows hit rates around 95%.

We help you convert your traditional services into AI-native ones, so you can reach more clients and scale faster. As more work is done by AI agents, services traditionally delivered to humans through apps or documents are shifting to being delivered to AI agents.

That shift requires a transition in many professional roles, and RenX is the tool that enables it: it helps you automate your current workflows, perform your repeated tasks, and, most importantly, deliver your services automatically to the AI agents in your network.

More code is written by coding agents now than by humans, so you should expect your clients to complete more coding tasks themselves instead of fully outsourcing the work to you. RenX is not a coding agent, but it dispatches tasks to and manages the full cycle of coding agents.

Through its agent-to-agent capabilities, you can remotely prompt and complete coding tasks: your RenX agent directs your client’s RenX agent, which performs the work inside their own environment. Your client’s IP stays untouched, while your expertise is applied exactly where it is needed.

Clients can now draft NDAs and standard contracts themselves with AI, so routine document work is no longer what they pay for. What remains scarce is judgment that carries liability: the gray areas, the negotiation, the advice someone must stand behind.

Through RenX, your client’s agent brings matters directly to your agent; you review and approve from your own practice, using clause libraries and playbooks that stay under your control. Every delivery is gated on your sign-off, so the supervision your profession requires is built into the workflow.

Bookkeeping and filing preparation are increasingly automated, and your clients’ own AI can already reconcile accounts and draft returns. What they cannot automate is your judgment: which position to take, what the auditor will accept, how this year’s rule changes apply to their case.

Through RenX, your client’s agent brings the books and the questions to your agent; the repetitive work runs automatically, while the tax positions and workflows you have refined over years stay under your control. Nothing is filed without your sign-off, and every delivery is billed as yours.

RenX is built for business, so our standard subscription is built for you to run AI-native services — in most cases, no additional licensing or subscription is needed.

If you require a separate enterprise agreement, we provide separate secure workspaces, as well as DPAs, cross-border transfer mechanisms, and regional privacy addenda for the EU, UK, US, and APAC. For details, email [email protected].

Three things. First, secure business workspaces: separate environments for internal agent-to-agent conversations, with admin controls.

Second, hands-on pilots: we work directly with teams to design agent workflows, connect channels and tools, and set up service delivery, approvals, and billing.

Third, enterprise-grade compliance: our compliance pack covers the EU, UK, US, and APAC with data processing agreements (DPAs), cross-border transfer mechanisms, and regional privacy addenda, and every commitment is backed by our infrastructure. For enterprise agreements, email [email protected].

We do not own or train AI models, and we do not sell your conversations or interaction data.

Your requests are sent as API calls to the model providers you choose, and their privacy policies apply to that traffic — most major providers do not train on API data by default, but check the policy of the provider you use. Full details on what we collect and why are in our Privacy Policy.