Super Chain works in five steps. You connect the firm’s tools, such as Microsoft 365 or ClickUp, and load its documents. You ask a question in plain English. It prepares the work, showing each step and citing sources. You approve any action set to need approval. A saved Skill lets the team repeat it.
Connect, Ask, It prepares, You approve, It repeats. The firm sets the rules once; after that, each request follows the same path, and a person stays in charge of anything that changes a record outside Super Chain.
The demo uses invented companies and people. Every screen in it is copied from the product.
Connect the apps your workspace uses, so the assistant can work in them as you.
Work the org's ClickUp workspace from chat: read tasks, owners and due dates, and create, update or comment on a task.
Choose when the assistant may use each tool for you. "Needs approval" asks you in the chat first.
A connector links Super Chain to a system the firm already uses, such as Microsoft 365, ClickUp or HubSpot. Each person connects their own account where the vendor allows it, so the agent acts with that person’s access, not a shared admin login.
Then each tool inside a connector gets one of three settings:
The firm’s admin. For each connector, the admin can give a person, a group or a role Read only, Read and write, a hand-picked set of tools, or no access. A tool someone may not use is not offered to the agent for them at all.
A person can tighten their own settings further, but never loosen past the workspace rule. Documents follow the same idea: each folder is shared at a level, and the agent only searches what the person asking may read.
For the full access model, read AI access control for financial firms.
They type it in plain English, as they would to a colleague: “Which of our loan documents mention a change-of-control clause?” or “Draft a task for each overdue review.” Files can be attached to the message.
Next to the chat box sits a choice of Ask or Auto. Ask, the default, means every tool set to Needs approval waits for a click. Auto lets those tools run without the wait for that conversation. Neither mode can switch on a tool set to Never allow.
Before any model runs, the message is screened for attempts to trick the agent. A message that looks like one is refused and recorded.
It narrates as it goes. The chat shows each step in plain words, such as “Searched documents” or “Looked up assets”, with what each step found. For a larger job it hands parts to specialist subagents that work in parallel, each shown as its own card.
Answers cite the firm’s own documents, so a reviewer can open the source page. When figures come from a connected data source that supports it, Super Chain checks each figure in the answer against the stored results and removes any it cannot support.
Word, Excel and PowerPoint files it creates stay in the chat until a person saves them to the vault. When it edits a Word document, it keeps the earlier version and stores a redline of what changed.
When it reaches a tool set to Needs approval, it stops. The chat shows what it wants to do and the details it would send, with two buttons: Deny and Approve. Nothing runs until the person decides.
A second check sits behind the button. Before a tool that writes to another system runs, an independent check asks whether the action is destructive, whether it matches what the person asked for, and whether it reaches for secrets. If the answer is doubtful, the call is refused.
Every decision, allowed or refused, is written to an activity log the firm holds, with the conversation it came from.
It is saved as a Skill: a short written workflow, in plain steps, that the agent follows with the tools the person already has. A Skill holds instructions only; it carries no code of its own.
Anyone with access types / in the chat box, picks the Skill and runs it again. The same permissions and approvals apply every run. Skills do not run on a timer: a person starts each one.
On infrastructure the firm controls: its own servers or its private cloud, isolated from other organisations. The trust and security page sets out the architecture, and the on-premise AI appliance page covers running it on the firm’s own hardware.
This guide is for CIOs and COOs at investment and wealth firms, family offices and brokers in Hong Kong, the UK and worldwide, who want AI to do real work across their tools without losing sight of who approved what.
It is not a technical integration manual. Single sign-on, network design and model choice are agreed with each firm during deployment.
Super Chain is accountable for the platform and its configuration. The office remains accountable for which documents go in, who may see them, and the terms of the model provider it chooses.
Super Chain is built by Super Cat Technology. The same team built a fully on-premise retrieval system for a Big 4 firm, where every answer returns with its source paragraph.
In five steps. The firm connects its tools and loads its documents. A person asks a question in plain English. Super Chain prepares the work and shows each step. The person approves any action that needs approval. A proven workflow is saved as a Skill and run again from the / menu.
The model, the documents, the prompts and the audit log run on infrastructure the firm controls, isolated from other organisations. Staff use it like any chat assistant, but it answers from the firm’s own files, only shows each person what they may see, and records every action for review.
Each connector tool is set to Always allow, Needs approval or Never allow. When the agent wants a tool set to Needs approval, the chat shows what it plans to do and waits. The person clicks Approve or Deny. Nothing set to Never allow runs, in any mode.
Ask is the default. In Ask mode, every tool set to Needs approval waits for a click. In Auto mode, those tools run without the wait. Auto never widens access: tools set to Never allow, the workspace rules and who may use a connector are all decided first.
No. A Skill is a written workflow, not a timer. A person starts each run from the / menu in the chat box, and the same permissions and approvals apply every time. That keeps a named person behind every action the agent takes on the firm’s behalf.
A 30-minute call with our team. We look at where your firm uses AI today, which documents and systems it would touch, and where it should run. You leave with a plan: what to start with, what to keep out, and which deployment fits. It is free.
Choose a time below. If you would rather see the platform first, book a private demonstration.