One business.
One Twin.
Any AI.

Your team already uses Claude and ChatGPT. Your Twin connects them to the systems you run, so every answer, dashboard and workflow works from live records, not uploaded spreadsheets, and each person sees only what they are allowed to.

From £75 a user a month Enquire about UK hosting

Discuss your first workflow

Your Twin’s app

ClaudeYour Twin connected

What needs a decision before Friday?

Asking your Twin

Asked your Twin · 3 records from

Three decisions are due by Friday. Northgate relaunch needs a delivery date: the plan has 9 October, but Kestrel Freight asked on Monday to move it to the 16th. Lumen Print's bill INV-8812 for £6,480 falls due on 30 September, and the supplier has asked to split it across two purchase orders. The cleaning contract renews on 1 October, and Ashgrove Facilities has quoted 12 or 24 months.

Decision	Evidence	Owner
Northgate delivery date	Plan: 9 OctKestrel email, 21 Sept: 16 Oct	Ana Duarte
INV-8812 supplier query	Bill due 30 SeptLumen email, 22 Sept	Leah Carter
Cleaning contract term	Renews 1 OctAshgrove quote	Tom Reeve

From your Twin, with your permissions. Each record opens in your Twin.

Reply to Claude

ChatGPTYour Twin connected

What needs a decision before Friday?

Asking your Twin

Asked your Twin · 3 records from

Decision	Evidence	Owner
Northgate delivery date	Plan: 9 OctKestrel email, 21 Sept: 16 Oct	Ana Duarte
INV-8812 supplier query	Bill due 30 SeptLumen email, 22 Sept	Leah Carter
Cleaning contract term	Renews 1 OctAshgrove quote	Tom Reeve

From your Twin, with your permissions. Each record opens in your Twin.

Ask anything

CopilotYour Twin connected

What needs a decision before Friday?

Your Twin agent

Your Twin agent · 3 records from

Decision	Evidence	Owner
Northgate delivery date	Plan: 9 OctKestrel email, 21 Sept: 16 Oct	Ana Duarte
INV-8812 supplier query	Bill due 30 SeptLumen email, 22 Sept	Leah Carter
Cleaning contract term	Renews 1 OctAshgrove quote	Tom Reeve

From your Twin, with your permissions. Each record opens in your Twin.

Message Copilot

Answers, with their sources

Read-only evidence path

Your TwinOne governed picture of the business, built from your records.

Connect the systems you useKeep your original records in the systems you use

Accounting · payments

Xero

SunSystems

Stripe

GoCardless

Documents · files

Microsoft 365

SharePoint

Google Drive

Knowledge · CRM

Google Docs

Google Sheets

Notion

HubSpot

Email · calendar

Gmail

Calendar

Outlook

Chat · meetings

Slack

Teams

Granola

Engineering · data

GitHub

Linear

Postgres

MySQL

SQL Server

Other connections agreed for your first workflow

Mercury Labs has delivered for

Client work by Mercury Labs, the team behind your Twin.

Spend less time assembling the picture.

Your Twin has its own app, and it also answers inside the AI tools you already use. On their own, those tools search one system at a time.

In Claude, on its own
ClaudeOutlook connected
What needs a decision before Friday?
Searched Outlook · 2 emails
I found two emails this week that look like they need a decision. Kestrel Freight asked on Monday to move the Northgate delivery to the 16th, and Lumen Print asked to split an invoice across two purchase orders.
I can't see your project plan, your accounts or your contracts from here, so I can't tell you when either falls due, what the invoice is for, or whether anything else is waiting on you.
From your mailbox only. Other systems were not searched.
Reply to Claude
01Without your Twin
Claude, with its own connector to Outlook, finds two of the three decisions due by Friday, and none of their dates, amounts or owners.

In Claude, with your Twin connected

ClaudeYour Twin connected

What needs a decision before Friday?

Asking your Twin

Asked your Twin · 3 records from

Decision	Evidence	Owner
Northgate delivery date	Plan: 9 OctKestrel email, 21 Sept: 16 Oct	Ana Duarte
INV-8812 supplier query	Bill due 30 SeptLumen email, 22 Sept	Leah Carter
Cleaning contract term	Renews 1 OctAshgrove quote	Tom Reeve

From your Twin, with your permissions. Each record opens in your Twin.

Reply to Claude

02With your Twin

The same question to Claude with your Twin connected: all three decisions, with their dates, amounts, owners and the system each came from.

03A shared understanding
Your Twin links the same customer across systems and connects their invoices, projects and commitments. The sources stay available to inspect.
Twin chatShared context
Where are we with Northstar?
1 min ago
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
Retrieved with
Ask anything about this task…
ChatGPTVia MCP
Where are we with Northstar?
ChatGPT
Used your Twin
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
Xero
Linear
Gmail
ClaudeVia MCP
Where are we with Northstar?
Claude
Used your Twin
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
Xero
Linear
Gmail
GeminiVia MCP
Where are we with Northstar?
Gemini
Used your Twin
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
Xero
Linear
Gmail
Microsoft 365 CopilotTwin agent
Where are we with Northstar?
Copilot
Your Twin agent
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
Xero
Linear
Gmail
TerminalTwin CLI
$ twin query --question "Where are we with Northstar?"
ANSWER
The website is delivered. The £8,400 invoice is overdue, and the client expects to pay Friday.
SOURCES
01 Xero / INV-2048
02 Linear / Northstar website
03 Gmail / Latest client update
The same context, from your terminal.
One Twin. Your tools. The same business knowledge.Example interfaces · External tools connect with your permission
04A useful answer
Ask the same question in Twin chat, ChatGPT, Claude, Gemini or Copilot, or from your own code through the CLI, SDK and API. Each uses the same connected records and definitions from your Twin. In Microsoft 365 Copilot and Teams, your Twin appears as an agent your team can ask.

Build on it with the AI you already use.

If your team is already building with Claude or ChatGPT, keep going. Your Twin gives those tools live records from your systems, joined and permissioned, so nobody exports or uploads the data first.

In Claude, with your Twin connected

ClaudeYour Twin connected

Chart what’s overdue by customer.

Asked your Twin · overdue by customer

Here is what is overdue by customer, from the invoices your Twin read from Xero this morning.

Overdue by customer£9,223.08

Tarnwick University£6,187.50
Loxley & Venn Freight£2,245.00
Silverhook Chauffeurs£519.95
Orrin Street Business Services£270.63

Live from your Twin · 7 invoices · nothing uploaded

I can save this to your Finance channel as a dashboard, so the team sees it without asking.

Reply to Claude

01In Claude and ChatGPT
Connect your Twin once. Ask for a chart, a summary or a draft, and the assistant reads the figures from your records when you ask.
02In your own code
The CLI, SDK and API give your scripts and internal tools the same records and the same definitions.
03In your Twin
Save what works as a dashboard your team shares, or run it as a workflow on a schedule.

Connected records refresh on the schedule agreed for each source. Each person, and each tool they use, reads only what they are allowed to see.

Your Twin finds the match. You make the call.

When records in two systems look like the same customer, your Twin brings the match to your Inbox with the evidence side by side. Accept it, decline it or leave it for later. An accepted match can be undone.

Agree the meaning. Reuse the understanding.

Which records describe the same customer? What counts as overdue? Your Twin connects identities, relationships and definitions, then checks the calculations against your data.

Your Twin’s app · Sample workspace

Define it once. Ask it anywhere.

Ask three people what is overdue and you may get three numbers. Your Twin agrees each definition with you once, then uses it for checked answers wherever your team asks.

Xero invoice

INV-0027

Customer: Tarnwick University
Amount due: £6,187.50
Due date: 10 Sep 2026
Status: Awaiting payment

Definition v1Proposed Accepted

Overdue invoice

Money is still owed
The due date has passed
Not voided or deleted

On your data0 of 16 open invoices

Asked in Twin chat

“How much is overdue?”

£9,223.087 invoices, 4 customers

Used Overdue invoice v1

Asked in ChatGPT

“What are we owed that’s late?”

£9,223.087 invoices, 4 customers

Used Overdue invoice v1

Use this definition?LaterDeclineAccept

Acceptedby you, just now

01Drafted from your data
Your Twin reads the systems you connect and drafts each definition in plain terms: what it counts, and what it leaves out.
02Accepted by you
Each definition comes to your Inbox. Your Twin uses it for checked answers only once an admin has accepted it.
03Named in checked answers
A checked answer shows the definition and version it used. A change is a new version, back in your Inbox for review.

An answer you can check.

See who owes what, using the same definition of outstanding debt. Inspect the invoices, ageing calculation and sources behind the answer.

the sources are right here

Your Twin’s app · Sample workspace

The figures are calculated. Never guessed.

AI models are good with words and unreliable with sums. So your Twin’s graph engine does the arithmetic: it holds your connected records and agreed definitions in memory and works out each figure. The model plans the question and writes the answer round it.

OverdueCalculated by the graph engine

£9,223.08

7 invoices, 4 customers

Definition: Overdue invoice v1
Counted from: Xero sales invoices
Read with: The access of the person asking
AI model used for the sum: None

Sample figure. Open it to see each invoice behind the total.

Same question, same number: The engine follows your definitions and uses no AI model, so the figure is the same in your Twin’s app, Claude and ChatGPT until the records change.
Answers in milliseconds: It is built in Rust and keeps your organisation’s records in memory, so a question across several systems does not wait on each of them.
Every figure traces back: Each result comes with what was counted and how, down to the invoices, bookings or tasks behind it.
Each person’s own view: It reads only the records the person asking is allowed to see, so a total never includes a row they could not open.

Your Twin. Your infrastructure.

A dedicated server for your business on the Core, Scale and Enterprise plans, with control over where it runs. The Team plan runs on a shared deployment.

European Union
Available today
United States
Available today
United Kingdom
In development

The region is where your server runs. Model processing can take place elsewhere, depending on the provider you use.

Dedicated server: On Core, Scale and Enterprise, your Twin runs on a server dedicated to your organisation.
Read-only connections: It reads the systems you connect without changing the original records.
Your model provider: Requests go to the AI provider configured for your organisation.
Audit trail: Checked answers show the definition, version and sources they used.

Cyber security certifications

Mercury Labs, the company behind your Twin, holds both certifications through its legal entity, Mercury Digital Assets Limited. Scope: whole organisation.

Mercury Labs is a Crown Commercial Service supplier.

Legal & Security

Why keep AI in the UK

Your question never leaves the UK.

For many firms the hard question about AI isn’t what it can do. It’s where the data goes. A fully UK-hosted Twin keeps the whole path, from the question to the answer, inside the UK.

01Say yes to sensitive workClient files, contracts and personal data stay inside the line your policies draw, so AI can reach the work it was kept away from.
02No outside model providerPrompts and business context go to a private model on UK GPUs, not to a third-party AI service.
03One boundary to reviewServer, storage, model, logs and backups sit in one place, so your security review has one set of answers.
04The same checked answersSources, definitions and checks work as they do today, whatever model answers.

Runs on Project Venus, our private model: a draft model that predicts tokens ahead for speed, an adapter trained on real tasks, and no metering on token usage, subject to concurrency thresholds. Read the research

United KingdomIn development

ServerUK datacentre

StorageUK datacentre

Private modelUK GPUs

Logs and backupsIn the UK

Outside AI provider

Not on the route

Everything on the route, from the question to the answer, runs inside the UK.

Your choice of model

Or use the models you already know.

Your Twin can answer through Venus, our private model, or through OpenAI, Anthropic or Google. Whichever you choose, it works from the same records and definitions, and each checked answer shows its sources.

Requests go to the model configured for your organisation.

Your Twin’s app · Sample workspace

Model

Runs privately in the UK or EU

Forward-deployed engineering

Our engineers work alongside your team.

Reserve engineering time to connect your systems, check a first workflow or extend your Twin. We agree the work with you.

01Work with our engineersWork directly with engineers on your systems, business questions and the results you need.
02Faster path to live useConnect the systems for a useful workflow and check the results with your team.
03Expand where it helpsAgree additional connections and workflow improvements as your needs change.

£1,200 a day. Larger or dedicated teams are agreed separately. Add engineering days

Your Twin’s app · Sample workspace

We connect the systems. You help define the result.

Start with one important workflow. In two focused weeks, we connect its systems and check the results with your team.

Days 1–5
Connect priority systems
We agree the question, connect the systems it needs and start with read-only access.
Days 6–10
Build and check your Twin
Your Twin connects the records, defines the calculations and tests them against your data. Work that fails the checks is left out.
End of week 2
Review the Twin, then switch it on
Check the result against real business examples with your team. The release stays off until you switch it on.

Get started today on Team, from £75 a user a month on a shared deployment. Core from £3,000 a month with a dedicated, managed deployment and AI usage from £250 a month. Optional engineering at £1,200 a day. See pricing

The questions we hear first.

Keep your existing systems. We agree a first workflow and connect the systems it needs, starting with read-only access.
Mercury Labs engineers work with your team to connect the relevant records, agree business definitions and test the calculations against your data. You review the result before switching it on.
Keep using them, with your Twin connected. On their own they search your systems one at a time, so an answer depends on which system was searched and how the question was worded. Your Twin joins your business records and agreed definitions into one picture, so ChatGPT, Copilot and your team get the same answer, with the sources and calculations it used.
Keep your business knowledge in the Twin as your choice of AI tools changes. Each external connection needs the right setup and permissions.
We need to check your requirements with you. Your Twin starts with read-only connections and access limited to your organisation. You can review each tested version of the Twin before it is switched on.
The standard checked query path does not include browser, terminal or write access. Your team still needs to approve where data and model requests go, how long evidence is kept and how the deployment is operated.
It tells you when the checked Twin cannot answer. This path (called a governed answer) identifies the published definition, version and source freshness it used. If the question or evidence is unsupported, it returns “not supported”.
An agent can still explore the underlying records within its permissions. That exploratory work is not labelled a governed answer and needs to be reviewed accordingly.

Start with the question your team keeps chasing.

Show us a recurring report, an awkward handover or a decision that takes too long to prepare. We agree a first workflow with you, then connect its systems and check the results with your team in two focused weeks.

Discuss your first workflow Partner with Mercury Labs

Spend less time assembling the picture.

01Without your Twin

02With your Twin

03A shared understanding

04A useful answer

Build on it with the AI you already use.

01In Claude and ChatGPT

02In your own code

03In your Twin

Your Twin finds the match. You make the call.

Agree the meaning. Reuse the understanding.

Define it once. Ask it anywhere.

01Drafted from your data

02Accepted by you

03Named in checked answers

An answer you can check.

The figures are calculated. Never guessed.

Your Twin. Your infrastructure.

Your question never leaves the UK.

Or use the models you already know.

Our engineers work alongside your team.

We connect the systems. You help define the result.

Connect priority systems

Build and check your Twin

Review the Twin, then switch it on

The questions we hear first.

Start with the question your team keeps chasing.