Agent tools & Agent Compute

Agents do more than write. They can read your project, search it, run code, and use a shell — which is how Multilo turns “graph this CSV” or “extract the tables from this PDF” into a real result instead of a guess.

Reading & searching

Agents can read documents (including the text of PDFs and Office files), search the whole project, and edit your draft. This grounding is what lets them work from your actual sources. See Super Search.

Agent Compute

When no built-in tool fits, an agent can write and run code. Agent Compute runs Python in a sandboxed, off-thread WebAssembly worker that downloads on first use — light to install and safe by design — so any file can become any outcome: a chart from a CSV, text extracted from a PDF, a converted format, an analysis.

Native escalation

For heavier work that needs real packages, agents can escalate to a native runtime and install dependencies on demand. This path is gated by your trust and consent — it only happens when you allow it.

The integrated shell

Agents run shell commands in a real terminal — PowerShell on Windows — so they can drive tools and scripts as part of a task.