AI assistants

Your workspace, inside theassistant you already use

Varynt Work speaks the Model Context Protocol, so any assistant that supports it can read your projects and act on them: as the person who connected it, never beyond what they can already do, and never with personal details or money.

The whole integration is one address

https://mcp-work.varynt.in/mcp

No API key to request, nothing for us to switch on. Paste it into your assistant and sign in as yourself.

included on every planhosted with one address, or local from npm

[ 01 ] What it can do

Real tools, not a demo

  • Answer questions about the work

    “What is overdue on the Acme site?” Your assistant searches tasks by project, assignee, status and due date and answers from your real workspace.

  • Pick up your own day

    Your tasks grouped the way you would ask: overdue, due today, this week, upcoming.

  • Create tasks and comment

    Raise work from wherever the conversation happened, filed to the right project and person, attributed to you.

  • Summarise a project or a client

    Status, milestones, open work, approvals and files in one answer instead of five screens.

  • See where the team stands

    Active, overdue and due-soon counts per person, and what is waiting on a client's decision.

  • Log hours

    “Put two hours on the Acme rebuild.” Log time, start and stop a timer, read back your week and send it for sign-off.

[ 02 ] What it can never do

A ceiling below your own access

Most tools let an assistant see whatever the person who connected it can see. For personal data that is the wrong default. Varynt Work enforces a ceiling below the person's own access, in the API, for AI sessions only. An owner reads all of this in the product; their assistant does not.

  • Read anyone's email address, phone number or location, staff or client
  • Open your client contact list, or a single contact record
  • See a rate, a cost or a margin
  • Open your workspace settings or your list of members
  • Change an employee or client record
  • Reach a workspace the person who connected it does not belong to

[ 03 ] How to connect

Hosted: one address

Varynt Work speaks the Model Context Protocol, the open standard AI assistants use to reach real tools. Your assistant can then read your work and act on it — as you, with your permissions, and never beyond them. The address below is the whole integration: there is no API key to request and nothing for us to switch on at our end.

  1. 01

    Step 1

    Open your assistant's settings and find where it adds a connector, an integration or an MCP server.

  2. 02

    Step 2

    Choose to add a custom or remote one, and paste the address above as the server URL.

  3. 03

    Step 3

    Your assistant will send you to the Varynt Work sign-in page. Sign in as you normally would.

  4. 04

    Step 4

    Read the permission screen. It names the app, shows where it will send you back, and says in plain words what it will be able to do.

  5. 05

    Step 5

    Choose how much access to grant — read-only is a sensible place to start — and approve it.

  6. 06

    Step 6

    The tools appear in your conversation. You can disconnect at any time from this page.

Every assistant words these menus differently and moves them between releases, so the steps above are deliberately generic. The one thing that does not change is that the address is the only thing you have to supply.

[ 04 ] Or run it yourself

The same tools, on your own machine

Prefer to keep it on your own machine? The same tools run as a local connector, installed straight from npm. Nothing of yours passes through a service of ours — your assistant talks to the package, and the package talks to your workspace.

Add this to your assistant's MCP configuration

npx fetches @varynttools/work-mcp on first run. Node 18 or newer.

{
  "mcpServers": {
    "varynt-work": {
      "command": "npx",
      "args": ["-y", "@varynttools/work-mcp"],
      "env": {
        "VARYNT_WORK_URL": "https://app-work.varynt.in",
        "VARYNT_WORK_TOKEN": "paste-your-personal-access-token",
        "VARYNT_WORK_WORKSPACE": "Your Workspace Name"
      }
    }
  }
}

Or install it yourself

npm i -g @varynttools/work-mcp

This route uses a personal access token instead of a sign-in, so create one first under My profile → API tokens and keep it like a password. The workspace line is only needed when your account belongs to more than one.

@varynttools/work-mcp on npm

[ 05 ] Trusted access

What stays true whichever route you use

  • No key to hand out

    The hosted route uses OAuth against your own sign-in. Nobody pastes a long-lived secret into a chat window.

  • It can only ever be you

    An assistant acts with the permissions of the person who connected it, resolved fresh on every request.

  • Recorded under a real name

    Anything an assistant creates or changes is attributed to the person whose account it used, in the same activity log and audit trail.

  • Off in one click

    Every connection is listed in that person's profile with what it may do, and they can disconnect it at any time.

  • We keep no copy

    The connector passes requests through to your workspace and holds no store of its own.

  • Included on every plan

    No add-on and no per-seat charge. Whatever your assistant costs is between you and whoever makes it.

[ 06 ] Questions

Asked by security reviewers

Does Varynt Work have AI built in?
Not yet. Everything on this page is your own assistant reaching your workspace over MCP, which works today on every plan. AI features inside Work itself are not built; when they ship they will be priced separately from the subscription plans.
Which assistants can connect?
Any that implements the Model Context Protocol, including Claude, ChatGPT and Cursor. There is nothing assistant-specific on our side, so a new one works the day it adds MCP support.
Can one person connect an assistant to the whole company's data?
They reach exactly what they could by signing in themselves, minus the personal details, money and settings withheld from AI sessions.
Can we stop people connecting assistants at all?
Yes. Talk to us: it is an account-level setting rather than something one user switches on for the company.
What if we change our mind?
Disconnect it from My profile, Connected apps. Access ends immediately; anything it already did stays, under the name of the person who authorised it.
Does the assistant train on our data?
That is governed by the assistant maker's terms, and worth asking them. What we control is what leaves your workspace: your permissions, minus what is withheld from AI sessions entirely.