Nebius and Toloka to Introduce Integration to Bring Human Experts-on-Demand to AI Agents

While building AI agents is more accessible today than ever before, moving one from a successful demo into a reliable production environment remains a significant challenge. As workflows grow in complexity, developers inevitably hit a "reliability ceiling" — the point where edge cases, high-stakes decisions, and ambiguous data require more than just a better prompt.

Today, Nebius and Toloka are announcing plans to bring Tendem into the Nebius ecosystem. This integration further strengthens the Nebius AI stack, anchoring the raw intelligence of Token Factory and the autonomy of Tavily agentic search with a programmable layer of human reliability. Originally designed as the market's pioneer hybrid human-AI agent, Tendem is now the first platform to embed vetted human experts directly into agentic workflows — making expert judgment callable via the Model Context Protocol (MCP), the emerging standard for AI tool integration.

Once integrated into the Nebius stack, the platform will allow AI agents to escalate ambiguity to Tendem's network of 10,000+ verified experts across 20+ domains, treating human judgment as a high-latency, high-accuracy API call.

What is Tendem MCP

Tendem is the first platform to make human expert judgment a programmable reliability layer for AI agents. Built on Toloka's 10+ years of human intelligence infrastructure, Tendem transforms expert judgment into a callable, scalable layer inside your agent's workflow. It's the same core infrastructure behind the high-quality human training and evaluation data trusted by Anthropic, Shopify, and the world's leading AI labs.  

The problem it solves is specific: AI agents don't fail because they lack intelligence. They fail because they lack context and judgment. At enterprise scale, even a 0.1% error rate is unacceptable. A single hallucination in a pricing, compliance, or medical workflow isn't just a glitch; it is a liability that can cost millions. Tendem is the safety net.

When an agent hits a low-confidence threshold or a predefined policy rule, it triggers a Tendem call to a verified domain expert — the same vetted PhDs and specialists trusted to train the world's leading frontier models. Every output passes through AI-guided QA before landing back in your workflow. The result is a system-level approach to reliability, not a manual patch. In benchmarking, this architecture delivered 53% faster task completion compared to human-only baselines, with a 21.3% quality improvement over traditional freelance platforms on complex tasks. The largest gains came in completeness — the step-gate approach prevents the omissions and hallucinations where AI-only systems most commonly fail.

Nebius: The Complete Stack for Agent Builders

With the addition of Tendem MCP into its ecosystem, Nebius is architecting the complete infrastructure stack for production AI agents: intelligence through Token Factory high-performance managed inference; autonomy through Tavily agentic search; and now reliability through Tendem human verification. 

Assaf Elovic, Head of AI at Monday.com stresses: “Enterprise AI requires more than powerful models — it requires a cohesive ecosystem. Agents must be able to discover context, reason intelligently, and escalate to trusted human expertise when uncertainty arises. The integration of Toloka's MCP Tendem into the Nebius ecosystem, alongside agentic search and high-performance managed inference, represents a meaningful step toward building AI systems that are autonomous, accountable, and production-ready.”

Roman Chernin, co-founder and CBO of Nebius, added: "We want to provide a complete stack for AI builders. Nebius provides high-performance managed inference through Token Factory, and now, through Tendem, we're providing the reliability layer. Developers shouldn't have to manage multiple vendors to get human judgment — it should all be part of the same seamless ecosystem."

Harley Finkelstein, Shopify President, adds: "Trust is the currency of the agentic world - and Tendem is a trust multiplier. Making humans part of the architecture, not an afterthought, is a powerful solution when the stakes are real."

This new integration significantly de-risks AI adoption at the enterprise level — providing the governance and auditability required to deploy autonomous agents in high-stakes environments. Nebius is not simply providing infrastructure for models. It's building the full-stack environment required to design, deploy, and govern AI agents in the real world.

Olga Megorskaya, Founder and CEO of Toloka: "How work gets done is being fundamentally redefined. By bringing Tendem to the Nebius stack, we will make human expert judgment a callable, scalable resource for every agent builder."