tools

Kestro

A local AI assistant that can operate your browser, files, terminal and connected tools.

// category

AI assistant

home/kestro

// about

Kestro is a local/desktop AI assistant that works with the tools you already use: the browser, your files, the terminal and connected platforms like Telegram.

It keeps control where it belongs. Actions that modify anything ask for confirmation first, every action is written to an audit log, and you choose which model answers — from multiple AI providers down to fully local models.

Source — coming to GitHub // public release in progress — details land here when they exist

// capabilities

  • Browser interaction
  • File interaction
  • Terminal interaction
  • Telegram connection
  • Multi-provider models
  • Manual / automatic model selection
  • Action confirmations
  • Audit log

Operate real surfaces

  • Interact with the browser — navigate, read pages, fill workflows.
  • Read and edit files with your confirmation.
  • Run terminal commands and scripts under supervision.
  • Connect Telegram for a conversational remote surface.

Bring any model

  • Anthropic Claude, OpenAI GPT, Google Gemini, xAI Grok.
  • LM Studio and Ollama for fully local models.
  • Manual model selection when you want a specific one.
  • Automatic model selection for everyday use.

Controls & transparency

  • Confirmation prompts before modifying or destructive actions.
  • A persistent action/audit log of everything it does.
  • Runs on Windows, Linux and macOS.

// how it works

Kestro runs on your machine as a local assistant. Each capability — browser, files, terminal, Telegram — is a controlled surface the assistant can act on, with your permission.

A provider layer wraps the different model backends behind one interface, so Claude, GPT, Gemini, Grok, LM Studio and Ollama models are interchangeable. You pick the model manually, or let Kestro auto-select the best fit for the task.

The safety loop is structural: read actions run quietly, write and destructive actions pause for confirmation, and the audit log records what was asked and what was done.

Type
Local / desktop AI assistant
Platforms
Windows · Linux · macOS
Providers
Claude · GPT · Gemini · Grok · LM Studio · Ollama
Model selection
Manual + automatic
Safety
Confirmations + action audit log

// use cases

Guided file work

Ask Kestro to locate files, inspect code or apply well-defined edits — with a confirmation before anything is written.

Browser-assisted research

Have it read pages, summarize content and gather information as you keep an eye on what it opens.

Terminal help, supervised

Ask for commands or run scripts interactively while every execution is reviewed and logged.

Telegram-connected automation

Send tasks from Telegram and monitor results from anywhere, with the same confirmation and audit rules.

// limitations

  • Kestro operates the browser, your files, the terminal and connected tools. It is not a general controller of every native desktop application — don't expect it to drive application UIs it was not built to interact with.
  • Local model quality varies by hardware; cloud models are usually stronger but send data to their providers.
  • Agentic behavior is only as safe as its guardrails — confirmations and logs reduce risk, they don't eliminate it.

// responsible use

  • Review the audit log regularly. Treat the confirmation gate as a real control, not a formality.
  • Use local models for anything you don't want to leave your machine.

// related writing

all articles →

// more software

Explore the other tools

Build something with me