What it does

Everything your agent gets when it comes to work.

The systems it connects to, what it knows and remembers, where a person steps in, two examples from start to finish, and what stays with your company.

Connect what you run

It works with what you already have.

Keep the agent you already run and the tools your company already uses: the CMMS, the issue tracker, the wiki, the team chat. Uffici connects to them through connectors we build and curate, so nothing is rebuilt. If you have none of them, or would rather not connect them yet, the basics are built in.

Bring your own agent

Uffici is not another platform to rebuild on: it connects to the agent you already run and gives it an office, so you move over step by step instead of starting from zero.

Works with the agent you already run

Uffici does not replace your agent. Keep the one your team already uses, and it does its work here: in the conversations, on the board, with the systems and the record.

  • HermesSupported
  • OpenClawComing soon
  • Claude CodeComing soon
  • CodexComing soon

Plugs into what you already run

Communication, knowledge, storage, workflow, CRM, analytics: the stack a company already pays for, worked by its agents in place.

Communication

  • Slack
  • AgentMail
  • Gmail
  • Outlook
  • POP3 / SMTP
  • Telegram
  • WhatsApp

Knowledge base

  • Confluence
  • Notion
  • SharePoint

Storage

  • Google Drive
  • OneDrive

Workflow

  • Infraspeak
  • Jira
  • Trello

CRM

  • HubSpot
  • Kommo
  • Pipedrive
  • Salesforce

SEO

  • Google Analytics
  • Google Search Console
  • Microsoft Clarity

General

  • MCP Server
  • Webhook

Or start with what is in the box

A board, tables, files and a clock come in the box. Nothing to buy alongside, nothing to wire together.

  • Coordinates

    A board The floor

    Tasks in columns, moving in sequence. Agents work it like anyone else on the team, and it is where an agent stops and waits for a person.

    • Lanes and cards with the fields your work needs
    • Checklists and comments on every card
    • Field rules that keep each card complete
    • A history of every move, and an archive for what is done
  • Records

    Tables The filing cabinet

    Structured data with the schema you need: the records an operation actually keeps, not a note the model wrote and hopes to find again.

    • Eight field types, from text and dates to multi-select
    • Filter, sort, count and total
    • Update or clear rows by rule, not one at a time
    • CSV in and CSV out
  • Reads

    Files The in-tray

    The CSV and the documents the work runs on. The unstructured half of the work, which is most of it.

    • Folders the team and its agents share
    • Copy, move, append and archive
    • Zip and unzip a batch in one step
    • Read a file a page at a time, or search inside it
  • Acts unasked

    A clock The alarm

    Agents run on a schedule as well as on a prompt, so the work happens whether or not somebody opened the app that morning.

    • Once, or on a repeating schedule
    • In the time zone the work keeps
    • Switched off by a person in one click
    • Every run lands in a conversation you can read

Bring the accounts you have. Start on what is in the box. Either way, an agent is working on day one.

Every connector is one we build and maintain ourselves, reached through a catalog we curate, never through whatever server someone pastes into a settings box. That is narrower than “connect anything”, on purpose: it is how your agents never end up running instructions written by someone who does not work for you.

Context and memory

Fewer agents. Each one knows where it is.

The answer to more work is not more agents. It is a few agents that understand the company, the team and the job in front of them, and remember what they learned last time. Uffici keeps that understanding in one place, so every agent you bring, whichever harness it runs on, arrives informed and gets better with use.

  • It knows the place

    How your company works, what is true for this team or this client, what this job is about. The same agent does the right thing wherever it is working.

  • It remembers

    What it learns in one conversation is there in the next, so nobody explains the same thing twice.

  • Different harnesses, one team

    Keep the agent each team prefers. They share the same context and memory, see each other’s work in the same conversation, and pick up where the other left off.

  • A person stays in charge of it

    Anyone can see what it remembers and correct it. What it knows belongs to the company, not to the agent: change the agent, keep the knowledge.

A few agents that know the company do more than a fleet that each know a little.

A person in the middle

The work runs on its own, until it needs a person’s judgment.

Integration tools chain one step to the next and run to the end, right or wrong. Real work is not like that. A draft comes back for a look. A priority gets questioned. An exception needs someone who knows the customer. Uffici is built for exactly that moment.

  1. The agent stops and hands over

    To a named person, with everything it saw: the conversation, not a summary of it.

  2. The person decides

    Edits it, sends it back, or refuses it. The agent picks up from there, as many times as the work takes.

  3. The decision is kept

    What was decided and why goes on the record, so the next time is easier, and it is clear who made the call.

In practice

A request comes in, and the whole loop runs on the company’s own system.

Two recurring jobs from two very different businesses, each one repetitive, each one losing the decisions that matter. Follow either one from the first request to the last step.

What stays with you

Your operation, your data, your call.

Anyone can promise this in a sentence. These four are decisions already made in how the system is built, which means you can ask us to prove each one, and we would much rather you did that than take our word for it.

What the agents learnYours

“Who owns what the agent learned?”

You do. It is your company’s memory, not ours and not an employee’s.

An agent’s memory belongs to the organization that created it, not to whoever happened to be talking to it. People join, move teams and leave; what the operation knows stays where it is, and everyone gets the same answer from it. That is the difference between building an asset and renting one.

The keys to your systemsSealed

“Does the AI ever see my passwords and keys?”

No. Not once, not even briefly.

The credentials for your systems are held somewhere the model simply cannot reach. It can ask for an action to be carried out; it never gets the key to carry it out itself. And a tool cannot talk its way past that by describing itself as harmless: what it is allowed to do is decided on our side, not claimed on its.

The history of what was decidedAuditable

“If someone asks what the AI did, and why, can you answer?”

Yes. Every decision is written down as it happens, in a record nothing can quietly rewrite.

It is kept apart from the day-to-day data the agents work on, one record per subject, with a retention you can point at. So when a customer disputes a call, an auditor asks how a decision was reached, or a regulator wants the reasoning behind an automated one, it is already written down: under GDPR today, and under the EU AI Act on its own timetable. A company that starts recording the day the question arrives has nothing to show.

The data itselfSovereign

“Who owns the data, and who decides what happens to it?”

You own it and you decide. We hold it for you, and we do as you instruct.

Your organization is the controller of everything in its workspace; we are only the processor, acting on your instructions. We do not sell it, we do not use it for advertising, and we do not train models on it. It is held under GDPR, isolated from every other company, and that isolation is enforced in the database itself, not merely in application code. Delete your content whenever you decide; the rights to reach it, correct it and take a portable copy of it are written into our privacy policy, not just into this sentence.

Already running an agent? Bring it in.

Tell us what you run, and a person on our team will talk with you about connecting it to Uffici, step by step.