Our GitHub account is temporarily unavailable: links to the code, releases and issues will not open.

Report a bug or talk to us

AI agents

AI agents overview

LLM is the part of OmniGet where you work with AI agents: chat with them, give them tasks, run coding tools such as Claude Code or Codex on your projects, and watch what they cost. Everything runs on your computer with the connections you add; OmniGet has no AI account of its own.

Where it is

Click LLM in the sidebar. The tabs run across the top of the window, and some tabs have a second row of links under them.

On the right of the tab bar, a badge shows work in progress and the usage monitor shows the limits of your connected accounts (Connections, models and limits).

The tabs

Tab What it is for Second row
Threads One thread per coding task, run by Claude Code, Codex, OpenCode, an ACP agent or OmniGet’s own agent on a project folder. See Threads.
Work Background work: tasks, routines that repeat until a check passes, missions, model comparisons and video summaries. See Work. All work, Tasks, Routines, Missions, Arena, Summarize a video
Chat Talk to one of your agents, or to several at once in a room where you @mention the one you want. See below.
Agents Your team of agents, what they remember and the skills they can use. See Agents, memory and skills. Agents, Memory, Skills
Companies Groups of agents organised like a small company, with an org chart, issues, goals and costs. See Companies.
Catalog Ready-made agents, commands, skills and servers to install, and a setup assistant for a project. See Catalog and project setup. Catalog, Installed, Project setup
Tools The AI coding tools on your computer: install, update, sign in, diagnose. See Coding tools.
More Measurements and history: usage, sessions, retrospectives, reading and remote access. See Usage, sessions and retrospective. Usage and performance, Usage, Sessions, Retrospective, Reading, Remote access
Configure Opens Settings › AI and agents: the AI provider, connections, models and limits, and MCP servers. Connections, Models and limits, MCP

Chat

  1. Open LLM › Chat.
  2. Pick an agent in the rail on the left (Agents), or switch the rail to Chats to reopen an earlier conversation.
  3. Type your message. To start over, click New chat.
  4. To let the agent work in a project, switch the conversation chip from Personal to a project folder. After a turn that changed files, Undo puts the folder back to how it was before that turn (it needs git). Detach folder goes back to Personal.

To talk to several agents at once, click New room under Rooms, add agents, then @mention the one you want to answer.

What you need

  • For the agents in Chat, Work, Agents and Companies: at least one connection in Configure › Connections (an API key, a local model, or a subscription you sign in to with its provider).
  • For Threads: a coding tool such as Claude Code or Codex installed and signed in (Coding tools).
  • For summaries and the AI video tools: the AI provider.

Your use of each provider follows that provider’s own terms and limits. OmniGet shows those limits; it does not change them.

Use OmniGet from other AI apps

The same agents and tools are available outside the app: