Omnigent puts multiple AI agents under one shared control layer
Omnigent is an tool for running Claude Code, Codex, Pi, and s through one shared workflow. A session can start in a terminal and continue in a browser or on a phone, with messages, files, terminals, and kept in sync. Several agents can work in the same session, so one agent can review another agent's work or different agents can split a larger task.
Models can be chosen through direct API keys, Claude or ChatGPT , or gateways such as and Ollama. A deployed server lets teammates watch a live session, chat with the agent, co-drive the session, or fork the conversation and continue separately. Policy controls can require approval before risky actions, limit tools, cap token use, or set a hard dollar budget with warnings before the limit.
Agent work can also run in such as Modal or Daytona instead of relying on a laptop. s are defined with a short YAML file that lists the prompt, tools, and optional .
Key points
- Claude Code, Codex, Pi, and s can run in one shared session.
- Sessions can move between terminal, browser, and phone while staying in sync.
- Teams can watch, co-drive, or fork an agent session.
- Policies can limit shell commands, file edits, , token use, and dollar spend.
- s are described with a YAML file containing prompts, tools, and .