Understand any project
Open a folder and ask about its structure, dependencies, history, or current changes. Koodos reads the real workspace before answering.
Build securely with any model
on every machine
One agent runtime for your files, terminal, projects, and devices. Work locally or across your fleet—with you in control.
Install once, then open your assistant from any browser.
# Install Koodos for Windows
irm https://api.kprcli.com/dl/install.ps1 | iexWorks where your work already lives
One assistant across your local app, terminal, and device fleet. Pick a task and watch it run.
Koodos is a personal agent runtime. request ├─ local app ├─ terminal └─ remote device routing ├─ local Ollama ├─ OpenRouter └─ hosted gateway fleet ├─ desktop online ├─ laptop online └─ server paused
Open a folder and ask about its structure, dependencies, history, or current changes. Koodos reads the real workspace before answering.
Let the agent edit files, run the smallest relevant tests, inspect failures, and explain the result with a clear activity trail.
See your fleet, route a task to the machine that has the files, and manage devices without moving your workflow.
What Koodos does
Project-aware. Tool-using. Multi-model. Built around your machines.
Reads and changes real files, follows repository instructions, shows what it used, and verifies work with relevant checks.
Executes commands on the selected machine and handles builds, tests, diagnostics, and development servers.
Knows the active folder, project tree, Git state, documentation, and files that make the request meaningful.
Loads task-specific instructions for repeatable workflows, from design implementation to device control.
Use OpenRouter, OpenAI-compatible services, hosted Koodos inference, or local Ollama models.
Connect data sources, web search, Figma, browser workflows, and custom tools without rebuilding the agent.
Delegate bounded work to specialists, keep persistent workers available, and gather results in one conversation.
Install it on a desktop, laptop, or server and route work to the device with the right environment.
Bring your own key, use your own endpoint, connect a local model, or route through Koodos while keeping the workflow exactly the same.
Supported providers
+ any OpenAI-compatible API