Valo is a complete workspace where AI can act without becoming blind, uncontrollable, or unsafe. Claude and Codex understand the app itself; every action remains visible and steerable. Keyguard lets them use encrypted credentials without exposing secrets to the model, granular permissions control exactly what tools may do, and guards stop dangerous commands. Memory, browser control, and orchestration are part of the same coherent system. Native on Windows, macOS, and Linux. Free to use.

AI software has offered a false choice: autonomous agents that are powerful but difficult to observe and control, or visible chat interfaces that cannot safely act across a real workflow. Valo removes that tradeoff. Claude and Codex understand the workspace around them, while every action stays visible and steerable. Keyguard allows the AI to use encrypted credentials without those secrets ever entering the conversation. Granular permissions define exactly what each tool may do, and guards stop dangerous commands before they run. Memory, browser control, orchestration, and recovery belong to the same system instead of being separate products bolted together. I have directed Valo's development from the beginning and used it to build itself every day. The result is a workspace designed around trust: enough freedom for AI to do meaningful work, enough visibility for a person to remain in control, and security built into the boundary between them.

Honestly the whole visible-action approach is really cool and feels long overdue for AI tooling. One thing I'd love to see is a "dry run" toggle that lets you preview the full chain of tool calls before any actually execute, kind of like a confirmation summary you can step through. Would make it way easier to trust the agent on bigger tasks.

Valo

A true workspace for AI, not another chat window

A true workspace for AI, not another chat window