02 / IT · Analysis
Praxi-Bot sets out a design for approving model tool actions
Once a shell and browser are connected, desktop agent developers still have to decide which actions proposed by the model can run.

The Praxi-Bot developer outlined a design for explicit tool approval in a Dev.to post on 20 Nov 2026. The post addresses how developers of desktop agents control what those agents can run.
The post identifies the risk that actions proposed by a language model could run on the system without restriction.
Connecting tools leaves a decision for developers. They still have to determine which actions proposed by the model can run.
Approval after tools are connected
The developer gives a shell and a browser as examples of tools that can be connected to a model.
In the design the post describes, connecting a tool and allowing a specific action to run are separate decisions. Explicit approval governs the decision to run the action.
How the desktop app is built
The post describes Praxi-Bot as an open-source agent under the Apache-2.0 licence.
Runtime. According to the post, Praxi-Bot runs around a daemon written in Rust. Its desktop app is built with Tauri.
The developer identifies two tasks that remain. The first is deciding which actions require approval. The second is making those decisions visible.