What is Daslab?

Daslab is a workspace where AI agents do real work in the systems you already run. You give direction in plain language. The work happens in scenes you can look at, and anything that touches the real world waits for your approval.

The idea comes from the spreadsheet. A cell holds a value, a formula connects it to other cells, and when one thing changes everything downstream follows — no run button, no deploy. The limit was always what a cell could hold: numbers and text, while your actual business — the inbox, the codebase, the servers, the invoices — lived somewhere else, pasted in by hand and stale on arrival.

In Daslab, a cell looks at the real thing. The inbox, the repo, the draft email — each is an asset: governed, gated, audited. Cells show them, live, still connected to where they came from. Every workflow becomes a scene — its data, its tools, the agent that runs it, and the history of everything it did.

You ask, it drafts, you approve the diff

You ask. "Pull this week's open PRs, the related issues, and which ones blocked deploys. Write a one-page summary to the team Sheet."

The agent calls the right tools, joins the results, drafts the write, and shows you the exact diff. You approve. The Sheet updates. The run is recorded. Or it runs on a schedule — against your inbox, your deploys, your backlog — surfacing what needs attention without you asking.

Writes wait for you, and every run is recorded

Reads run freely. Writes stop at a gate: you see what will change before it does, and actions you trust can run on auto. The default is that you look first. Every run is recorded end to end — what the agent did, which assets changed, what it cost. When something looks wrong, you read the run instead of reconstructing it from memory.

100+ integrations, any MCP server, your own models

Gmail, Slack, Microsoft Teams, S/4HANA and SuccessFactors are among the native integrations; anything that speaks MCP plugs in alongside them. When no integration fits, the agent writes code and runs it in an isolated sandbox. The model is your choice — Anthropic, OpenAI, Amazon Bedrock, SAP AI Core, Gemini, OpenRouter — managed usage by default, your own keys on Enterprise. And you can reach your scenes from tools you already work in — Claude Code, Codex, Cursor, or any MCP client.

The long arc is physical

The thing we're really building is a runtime where conversations orchestrate any system that has a state — including physical ones. We have a hardware lab in Bangkok where we test this against real robots and microcontrollers. The same software that runs your accounting close should eventually run a chemistry experiment.

Today's product is one slice of that arc. Read more about where this is going.

Your first scene takes five minutes

For SAP teams specifically, start here.

Updated 2026-07-18