Connect Human Experts to Claude & ChatGPT via MCP

By Toloka Team · Updated June 2026

Your AI assistant is fast and capable … until it isn’t. Something comes along that requires more nuance than simply “getting things done fast”. You work on a task that needs checking against a real source, or something requires a judgment call where "sounds right" and "is right" aren't the same thing.

Until now, that meant leaving the chat for another tab, explaining the task over again before pasting the result back. But with Tendem, you get a Model Context Protocol (MCP) connector that lets your favorite AI tool hand work to a vetted human expert, bringing the finished result straight back into the same conversation. There’s no need to tab switch or copy and paste – it’s all there inside ChatGPT, Claude, and Cursor.

Here,  we look at what Tendem works, and how to get the most out of your tasks.

How Tendem MCP works

Tendem connects the AI tools you already use to a network of 10,000+ vetted human experts. When your agent encounters work that needs human judgment, you simply mention Tendem in chat (just type or say "Use Tendem to…”) and your agent picks it up automatically, the same way it would call any other connector.

That’s not to say Tendem is "AI doing your work for you." Instead, AI handles the busywork of scoping and routing the task, with a real human expert delivering the result. Which is the whole point, because an AI alone still hallucinates and falls short on complex, judgment-heavy work, and the cost of being wrong on business data costs you. When getting it wrong isn't an option, a person does it.

Tendem is built on Toloka's 10-plus years of human-AI collaboration infrastructure and is backed by Bezos Expeditions. When we say vetted human experts, it's a claim we can stand behind.

MCP, in plain English

MCP stands for Model Context Protocol  and is an open standard introduced by Anthropic in late 2024 that lets AI agents connect to outside tools and services. Think of it as a universal adapter, the USB-C port for AI agents. A service builds one connector, and it works across many AI clients.

In practice, it means your AI tool can call Tendem directly inside the conversation. You describe what you need in plain language, and the agent does the rest.

What works with Tendem

Tendem works inside Claude Desktop app, Claude.ai (web), ClaudeCode, Claude Cowork, ChatGPT, Cursor, OpenClaw, Codex, Hermes — practically any AI tool that supports custom MCP connectors. Of all the options, Claude is the most thoroughly tested experience as of writing this article.

How to connect Tendem

Tendem account: Sign up at agent.tendem.ai (free to start, no credit card required).

MCP clients we support: You can find step-by-step instructions for every supported platform on the Tendem MCP connection page.

From there, just mention Tendem when you describe a task.

How Tendem MCP works, from start to finish

Once connected, the entire workflow lives inside your chat. If you’re wondering what that might look like when you’re using it, it goes a little something like this?

Describe the task

Mention Tendem and explain what you need, the same way you'd write any prompt. Your agent may ask a few clarifying questions to make sure the brief’s on point.

Get a price

Tendem reviews and prices the task  (which takes up to 10 minutes) and the proposal appears right in the chat.

Approve in chat

Nothing is charged before you approve. If the price feels a little on the high side, ask your agent to narrow the scope and resubmit, or simply walk away. There’s no commitment.

An expert does the work

A vetted human expert picks up the task. Say a job needs multiple specialists, like a designer and a copywriter reviewing the same pitch deck, the agent splits it into parallel tasks automatically.

Results come back to the same conversation

You receive the output as readable text plus downloadable files, with a link to the full result in Tendem. And if you want to check on how it’s going, ask for a status update at any point with a simple "check my Tendem task."

More than 10,000 experts, across a range of tasks

If you want the best results, aim for specific tasks with a clear deliverable. Here are a few prompts you can try out for yourself to see Tendem in action.

• "Use Tendem to find me placements to promote [our new feature]. Give me 40 creators with newsletters, X, Youtube and podcasts in the USA and EU, 5K-30K audience size. Topics: [AI tooling, productivity, design, entrepreneurship]. Host profile [AI/ML expert, techie, educator]."

• "Use Tendem to clean and enrich this investor list – remove duplicates, fill in missing LinkedIn URLs, flag anyone who invested in AI startups in the last 12 months."

• "Use Tendem to write 5 LinkedIn posts from the key points in this doc. Professional but conversational, under 200 words each, clear CTA. Also add a list of 20 LinkedIn posts on the same topic [in this area], shared within the last 7 days, minimum 100 likes, good for commenting and visibility.”

• "Use Tendem to review this pitch deck. I want a designer's eyes on the layout and visual hierarchy, and a copywriter checking the messaging is on point. Call out what doesn't work and suggest fixes."

That last one needs two different people, which Tendem takes care of without you needing to do anything. Your agent reads it as a design job and a copy job, opens two tasks, and runs them in parallel. Both come back to the same conversation when they're done.

How the quality pipeline works

Here's the path your task takes before the result comes back:

Task intake and clarification

Tendem's AI agent reads the request and breaks it into subtasks, working out what information it needs along the way.

Expert matching and execution

With over 10,000 domain experts, tasks go to people best-suited for your needs. If it’s high-quality writing, then an experienced copywriter is involved. Need 500 prospects researched with their contact details verified? A research analyst takes it on.

Automated QA

Before it goes any further, AI runs an automated check over the work for quality and completeness.

Human QA review

A separate QA specialist reviews the output against the original requirements to make sure that everything lines up.

Structured delivery

The final result comes back as structured JSON. Inside you get the verified deliverable along with its source citations, quality score, and other metadata. Nothing is hidden, so you can see where the AI did the work and where the expert took over.

Tendem vs other alternatives

Here's how Tendem stacks up against the alternatives.

Versus doing it all in your AI agent

AI alone is fast and inexpensive, but it falls short on complex, long-spec work that’s judgment-heavy. Not only that, but it can produce output that sounds right without being right. In other words, the classic AI hallucination Tendem gives it to a person, right there in the chat.

Versus freelance marketplaces

You may be asking why you can’t just go straight to a traditional marketplace, like Upwork or Fiverr, and get everything you need done there. And rather than leaving you to find and brief someone yourself, Tendem does the scoping and routing, with domain experts who've spent years in their field.

On Upwork or Fiverr, you manage everything — briefing, back-and-forth, quality control. With Tendem, scoping and routing are handled for you. You stay in your agent and receive a finished deliverable.

Tips for getting the most out of Tendem

Like any prompt, you want to be specific when getting the most out of Tendem. Here are some tips for the best results.

Define escalation rules

Don’t rely solely on the agent’s judgment. Set explicit rules, like always getting financial data verified before you act on it.

Be specific

Provide enough context for someone to understand the job, then be clear about the level of accuracy expected and how the finished work should come back. The more the request feels like a proper deliverable, the better the result.

Keep working in parallel

You don't have to wait for one task to come back before starting a new one Carry on with other tasks in your chat and check in on your Tendem work whenever you want.

Frequently asked questions

How long does a task on Tendem take?

Pricing takes up to 10 minutes. Expert work takes real time depending on the task. We promise reliable, not instant. Most tasks complete within hours. You can keep working in the same chat and ask for a status update at any point.

What kinds of work do Tendem experts handle?

Tendem’s network spans 30+ domains including data research and analysis, finance, legal, technology, cybersecurity, content and copywriting, marketing, engineering, and scientific fields.

Can I use Tendem with non-Anthropic models?

Yes. Tendem is an MCP connector, and MCP is model-agnostic. Any AI application that supports custom connectors can connect to Tendem, including ChatGPT.

Do I pay before I see a price?

No. You see a transparent price in the chat before any work begins, and nothing is charged until you approve it. Cancel before approving and there's no charge.

How much does it cost to start?

New users get a $50 starting bonus on signup, and the first three tasks are 50% off.

Give your AI agent a human expert. Connect Tendem to Claude, ChatGPT, Cursor, or your AI tool of choice — no SDK, no code changes — and delegate without losing your context.

Connect Tendem →