Get started with Colossus¶
Colossus helps AI agents do work beyond a chat: inspect a repository, use tools, and carry a task across sessions. It makes their access, approvals, execution boundary, and audit evidence visible. Start with a local run, then connect the model and services you want to use.
Choose your starting point¶
-
CLI and terminal UI
Install the native command for scripts and interactive terminal work. The first run works offline without a model account or API key.
-
Desktop
Choose a folder and work in the native app. Desktop manages its own local runtime and includes the CLI.
Why Colossus exists¶
An agent becomes useful when it can read files, run commands, or reach a service. Those actions also need clear authority and a record you can inspect. Colossus brings the tools and the controls into one runtime for both connected and offline work.
-
Know what it can do
Access selects tools. Policy and approval govern each effect. The execution boundary controls which host resources the effect can reach.
-
Keep long work moving
Sessions, plans, and goals give multi-step work a durable shape. Return to a task after an interruption and inspect what remains.
-
Check what happened
Colossus records requests, decisions, effects, and uncertain outcomes in a verifiable journal so a restart does not silently repeat an action.
-
Work online or offline
Begin with a credential-free local proof, connect hosted or local models, and prepare controlled environments without changing the authorization path.
From first run to real work¶
The first run roadmap shows the path from an offline check to a connected model and a repository task. Read core concepts before widening access. When you are ready to explore beyond the first task, see Use Colossus.
Colossus is alpha software. Review upgrade and compatibility guidance before updating an installation you depend on.