Cloud.ru has released the public Agents Space platform, which allows users without technical skills to launch autonomous AI agents through a chat interface. The platform combines a catalog of Gemini, Claude, Qwen, and Kimi models with connectors to email, calendar, and Yandex.Metrica in a unified no-code environment with pay-as-you-go billing.

image
image
image

What happened

Cloud.ru launched Agents Space — a platform for working with personal AI agents without the need to configure models or servers. At launch, three agents are available: GigaAgent, described as an autonomous universal agent based on the Ouroboros project; OpenClaw for connecting to external services; and NemoClaw, which operates in an isolated environment. Users choose a language model from a catalog: Gemini 3 Flash Preview, Qwen3.6-35B-A3B, Kimi K2.5, Claude Opus 4.8, and Claude Sonnet 4.6. Billing is pay-as-you-go for vCPU, RAM, and tokens. New users are provided with a starting grant of 4,000 rubles. The platform is available at agents-space.cloud.ru.

Context

The platform is built on Cloud.ru's AI Factory and Evolution Foundation Models. Architecturally, Agents Space is a multi-model router over existing LLMs, not a new model or architecture — it is a no-code layer that lowers the barrier to entry for users without machine learning expertise. The platform provides connectors to email, calendar, cloud storage, and Yandex.Metrica, as well as a set of skills: marketing, text generation, and Managed RAG. The UX pattern: the user sets a task via chat, the agent independently plans the steps, connects tools, and returns the result.

Why this matters for the industry

This is the first public launch of a no-code AI agent platform from Russia's largest cloud provider. The product positions Cloud.ru as an infrastructure player in the AI agency segment and shifts the market from chatbots to platforms where AI performs tasks autonomously. For startups and product teams, the infrastructure cost of launching an agent solution drops sharply: instead of months of configuring an LLM backend and agent loop, you can now prototype in 15 minutes. However, dependence on the Cloud.ru platform creates a vendor lock-in risk. The pay-as-you-go model forms a natural funnel from test usage to commercial workloads.

Why this matters for users

The platform allows you to launch a working AI agent with RAG and connectors in 15 minutes without writing code. Users can test the platform for free with a starting grant of 4,000 rubles via agents-space.cloud.ru. Product teams can immediately build prototypes of agent workflows: marketing automation, RAG overlays on their own data, integration via built-in connectors. Agents work through an intuitive interface — the user formulates a task, and the agent independently plans and executes steps, connecting the necessary tools.

What is still unknown / limitations

The platform does not represent a new model or architecture — it is a wrapper over existing LLMs without a demonstration of scientific novelty. GigaAgent is described as an autonomous agent based on Ouroboros, but there are no publications, benchmarks, or open code confirming its technical characteristics. No results on standard agent benchmarks are presented: AgentBench, WebArena, SWE-bench. There is no documentation on API, rate limits, and SLA for latencies, which makes it impossible to use the platform in a production environment at the current stage.

Sources

Author

Look at AI, editorial team