Your own intelligence.
Agents,
Scout researches. Quill writes. Vera checks the work.
Any model.
Local through Ollama, or Claude and Codex with your own login or key.
A guard for
The Sieve guard filters suspicious instructions from tool results for local and OpenAI agents.
One line
Your whole office goes to work, and reports back.
Each agent has a role, a brain and a personality of its own, and sits at its own desk in your office.
Local models through Ollama, or Claude and Codex through your existing logins. OpenAI is also supported with an API key.
Drag agents, connectors, conditions and loops into a flow. Every run plays out live in the office.
Filters suspicious instructions from tool results for local and OpenAI agents. Claude-native tools run outside this filter.
With approvals enabled, shell commands, file writes and non-GET requests pause for Allow or Deny.
Pair Telegram with a code and run your systems, check on them or stop them from your phone.