Human-Overseen AI

You Decide
How Human Your AI Is.

Drop us into your app like any LLM. Then dial human oversight from 0% (fully autonomous agents) to 100% (every word written by a real person). Same API. Your call.

Get API Access See the Mix

OpenAI Compatible

Swap the endpoint. Keep your code. Our API speaks the exact same format.

Configurable Oversight

One slider, 0–100%. Run agents autonomously, sample for QA, or hand-write every reply.

Expert Routing

Legal query? Lawyer reviews. Medical? Clinician. The right human in the loop, when it matters.

Pick Your Mix.

Set human oversight per request, per agent, or per app. Pass human_oversight: 0.25 in your API call. We handle the rest.

Fully Autonomous

Pure AI agents. Fastest, cheapest. Use for low-stakes, high-volume work.

Spot Checks

Humans audit a random sample. Catch drift early without slowing things down.

Human in the Loop

Every high-stakes reply is reviewed before it ships. Bots draft, humans approve.

Human Written

Every word from a real person. Zero hallucinations. The original Human API.

Mix and match: route legal queries at 100%, marketing copy at 25%, internal tools at 0%. All from one endpoint.

How It Works

Four steps. That's it.

Send a Request

POST to our endpoint with your human_oversight level. Same OpenAI format.

AI Drafts (or Skips)

Our agent generates a response. At 100% oversight, we skip straight to a human.

Humans Review

Per your setting, a qualified person audits, edits, or rewrites the reply from scratch.

You Get JSON Back

Same response format as any LLM. Your code doesn't know the difference.

Drop It In.
It Just Works.

If your code talks to an LLM API, it can talk to ours. No new SDKs. No migration.

OpenAI SDK Compatible

Point the Python or Node SDK at our endpoint. Done.

Streaming Support

Real-time SSE streaming as humans type. Same format as OpenAI.

Flexible SLAs

Minutes for standard. Seconds for pre-briefed expert pools.

Diagram showing API request flow from your app through ManchesterHumans to human experts

Pricing

Two tiers. Transparent pricing. No hidden fees.

Standard

Pay for the Mix You Pick

Output cost scales linearly with your oversight %. Less human, less cost.

Input tokens £0.05 / 1K
Output @ 0% oversight £0.02 / 1K
Output @ 50% oversight £0.085 / 1K
Output @ 100% oversight £0.15 / 1K
  • ✓ Set human_oversight per request
  • ✓ 24-hour SLA on reviewed requests
  • ✓ Self-serve signup, demo key included
Get Started
Professional

Expert Tier

Custom rates Based on domain
Response time 1 min – 24 hrs
  • ✓ Vetted domain experts review at high oversight %
  • ✓ Legal, Medical, Financial, Technical, Academic
  • ✓ Per-route oversight policies (e.g. 100% legal, 25% routine)
  • ✓ SLA contracts with guarantees

API Reference

OpenAI-compatible. Drop-in replacement. One endpoint.

POST /v1/chat/completions

Send requests in the exact same format as OpenAI. Add human_oversight (0–1) to set your mix. Use your demo API key from signup.

curl -X POST https://manchesterhumans.com/v1/chat/completions \
  -H "Authorization: Bearer manchesterhumans-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "human",
    "human_oversight": 0.5,
    "messages": [
      {"role": "user", "content": "What are the key principles of contract law?"}
    ]
  }'
from openai import OpenAI

client = OpenAI(
    api_key="manchesterhumans-your-key-here",
    base_url="https://manchesterhumans.com/v1"
)

response = client.chat.completions.create(
    model="human",
    extra_body={"human_oversight": 0.5},
    messages=[
        {"role": "user", "content": "What are the key principles of contract law?"}
    ]
)

print(response.choices[0].message.content)
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "manchesterhumans-your-key-here",
  baseURL: "https://manchesterhumans.com/v1"
});

const response = await client.chat.completions.create({
  model: "human",
  human_oversight: 0.5,
  messages: [
    { role: "user", content: "What are the key principles of contract law?" }
  ]
});

console.log(response.choices[0].message.content);
// composer require openai-php/client
use OpenAI;

$client = OpenAI::factory()
    ->withApiKey('manchesterhumans-your-key-here')
    ->withBaseUri('https://manchesterhumans.com/v1')
    ->make();

$response = $client->chat()->create([
    'model' => 'human',
    'human_oversight' => 0.5,
    'messages' => [
        ['role' => 'user', 'content' => 'What are the key principles of contract law?']
    ]
]);

echo $response->choices[0]->message->content;
# gem install ruby-openai
require "openai"

client = OpenAI::Client.new(
  access_token: "manchesterhumans-your-key-here",
  uri_base: "https://manchesterhumans.com/v1"
)

response = client.chat(
  parameters: {
    model: "human",
    human_oversight: 0.5,
    messages: [
      { role: "user", content: "What are the key principles of contract law?" }
    ]
  }
)

puts response.dig("choices", 0, "message", "content")

Response format (standard OpenAI chat completion):

{
  "id": "mh-a1b2c3d4e5f6...",
  "object": "chat.completion",
  "model": "human",
  "choices": [{
    "index": 0,
    "message": { "role": "assistant", "content": "..." },
    "finish_reason": "stop"
  }],
  "usage": { "prompt_tokens": 24, "completion_tokens": 156, "total_tokens": 180 }
}

Try It Right Now

Real human-in-the-loop tasks. Content moderation, accuracy checks, query routing. Pick one or type your own.

Live Demo — Human API Ready
Pick a scenario above or type your own message below.

5 free demo messages per hour · No sign-up required

Need Help Shipping It?

Beyond the API. We scope, build, and run agent systems for teams that want human-in-the-loop baked in from day one.

Scope & POC

Discovery workshop, then a 2-week proof of concept with our oversight API wired in. We map oversight tiers to your real risk surface.

Agents, Skills & Intelligent Code

Tool-using agents, skill routing, retrieval, structured outputs. Native to the OpenAI SDK, native to our HITL endpoint. No glue layer.

Plan & Execute

Phased rollout, oversight policy, eval harness, expert pool setup, handover. From kick-off through to production with a steady-state SLA.

HITL, out of the box.

Every system we build hits the same human_oversight parameter you'd use yourself. Same SDK, same endpoint. Start at 100% oversight while the agent learns your domain — slide it down as evals say it's safe.

Book a Discovery Call

We also passionately believe AI should be built responsibly.

So we made a free open-source toolkit — starting with an AI agent that helps you get set up and going.

Free Ethical AI Toolkit

Checklists, templates, frameworks. All free. All practical. No sign-up.

Ethical AI Agent

An AI assistant that walks you through setting up responsible AI practices. It audits your setup, recommends frameworks, generates documentation, and helps you implement every part of the toolkit. Just talk to it.

Free · Open Source · Try It Now →

Bias Audit Checklist

Step-by-step bias identification and mitigation for training data, outputs, and deployment.

PDF + Interactive

Model Card Templates

Document model capabilities, limitations, training data, and ethical considerations.

Markdown + Docs

Impact Assessment

Evaluate social, economic, and environmental impact before you deploy.

PDF + Spreadsheet

Transparency Protocol

How to tell users, regulators, and communities what your AI actually does.

Documentation Kit

Data Governance Playbook

Responsible data collection, consent, anonymisation. GDPR-aligned.

Playbook + Templates

Incident Response Plan

When AI goes wrong, you need a plan. This is that plan.

Response Template
Get the Toolkit

No email gate. No sign-up. Just use it.

Get Your Demo API Key

Sign up to get a free demo key with 20 requests/day. Try the Human API in minutes.

Get the Free Toolkit

Everything you need to build AI responsibly. Checklists, templates, frameworks.

Get the Toolkit