Dashboard

One console for AI across your company.

In the zwrm dashboard you create agents, decide what each one may access, connect your tools and put routines on a schedule. Every run is logged with the steps the agent took and the tools it used.

Give each agent one job. And only the access it needs.

Create an agent, describe its job in plain words and attach the tools it may use. It works in its own isolated machine and never borrows your accounts.

  • A job in plain words

    The instructions go with every run, chat and schedule of the agent. Change them whenever the job changes.

  • The runtime you choose

    Claude Code or Codex with the subscription you already have, or the zwrm runtime with open models hosted in the EU.

  • Only the tools you attach

    An agent sees only the integrations attached to it. Set one to “Ask first”, and a run waits for your OK before using it.

  • Limits per day

    Cap the runs and the spend of each agent per day. Once it reaches a cap, the agent starts nothing new until the next day.

Connect your tools in one place and decide who uses them.

Add Microsoft 365, HubSpot, Notion, GitHub and other tools for the whole workspace. Then attach them to the agents that need them, with only the tools each one should have.

Workspace

Connected once

Add an integration once for the whole workspace. Its credentials stay in the zwrm gateway and never enter an agent’s machine.

Identity

Each agent signs in as itself

Authorize an agent with an account of its own, and every tool shows who did what. An agent never acts with your login.

Scope

Down to single tools

Pick which tools of an integration an agent gets, read-only if that is all it needs. What it was not given, it cannot call.

Notion, Microsoft 365, Microsoft Teams, HubSpot, Attio, DATEV, Front, GitHub, Slack, Linear, Asana, monday.com, Stripe, Canva, Sentry, PostHog, and any other tool with an MCP server.

All integrations

Routines start on a schedule or the moment something happens.

Describe the task in plain words and choose when it runs: every Monday at 8, each morning before the team starts, or whenever your shop, GitHub or a Teams chat sends an event.

Nothing piles up

If the last run of a schedule is still going when the next one is due, that one is skipped and logged instead of stacking up.

Picks up where it left off

A routine can keep its workspace, so this Monday’s report builds on last Monday’s files and notes.

Started from Teams

Connect a Teams chat, and a new message there starts a run. The agent replies in the same chat.

Choose the model for each agent. Or keep it all on EU-hosted models.

Run an agent on Claude Code or Codex with the subscription you already have, or on the zwrm runtime with open models hosted in the EU and paid from prepaid credits.

  • Claude Code

    Runs on your Anthropic account or Claude subscription.

    • Claude Opus
    • Claude Sonnet
  • Codex

    Runs on your OpenAI account or ChatGPT subscription.

    • GPT Sol
    • GPT Terra
    • GPT Luna
  • zwrm runtime

    Open models hosted in the EU.

    • GLM-5.3
    • Mistral Large 4
    • Kimi K3
    • DeepSeek V4 Pro
    • Qwen3 235B
    • and more

Every run is logged. Each step, each tool, each result.

Open a run to see what started it, what the agent did, which tools it called and what came out. Files it produced stay with the run after its machine is gone.

Also in the log

  • Every agent’s runs and chats in one list, whatever started them
  • Each tool call, with the account the agent acted as
  • Skipped schedules and refused deliveries, with the reason
  • Live runs to follow, and paused steps to approve or deny

Assistant · ⌘K

Mira Novak
What did the finance agent do on Friday?
Assistant
It ran open-invoices at 14:00. After your OK it sent reminders to Kestrel GmbH and Aster AB, and it saved an overview of the open invoices as an Excel file.

Tell us what lands on your desk every week. We’ll show you an agent doing it.

A demo with our team, in English or German. Bring one recurring task and the tools it touches.