Octomind Cloud organizes ongoing work into named octos
Octomind redesigns its agent workspace with dedicated folders and activity views while retaining shared machines and existing cloud tools.
Octomind Cloud has reorganized its agent experience around octos, named assistants assigned to individual workstreams. Its October 1 announcement introduces dedicated working folders and a dashboard organized around each assistant.
Each octo keeps its conversations, routines and memory together. Its conversations and scheduled routines run from its assigned working directory. The dashboard shows its activity, next scheduled action and monthly spending. Jobs can be delegated to specialist agents, with their results returned to the same conversation. Existing machines become octos while retaining their previous work.
What a separate folder means
Octomind says agents sharing a machine still use the same container and operating system user. A dedicated folder organizes files and context without creating a security boundary. The company recommends a separate machine when a job needs stronger isolation.
That distinction matters for anyone handling several clients. Organizing assignments under different names can make a workspace easier to follow, but it should not be treated as proof that one assignment cannot access another.
The existing tools remain underneath
This redesign builds on an existing cloud service. The product documentation already describes agents that browse, run code and write files, with scheduled routines and delivery through Telegram, WhatsApp or Slack. Machines provide a shell, sudo and Docker, and files remain available between sessions. These capabilities explain what the new organization sits on top of.
The underlying Octomind runtime is available under Apache 2.0 and can run locally. Its documentation describes configurable specialist roles, support for multiple model providers and switching models during a session. Cloud customers therefore have a hosted interface, while developers can inspect the runtime and use their own supported provider configuration.
The repository also documents rules that can block matching tool calls before execution. Those configurable controls are distinct from the folder arrangement. Their presence does not establish that a particular deployment has been independently audited or that every task will complete correctly.
How the costs add up
The current pricing page lists Pro at $50 per month, with a $20 introductory first month, and Max at $200 per month. Their monthly usage allowances are $50 and $200 respectively. Model usage is billed at provider prices plus 5 percent.
The allowance also pays for machine time, storage and generated media. Prepaid credits can cover usage beyond it, so the subscription price should not be read as unlimited model access.
The practical decision is whether grouping ongoing work by assistant makes supervision easier. Buyers still need to consider model consumption, shared machine access and how they will review completed work.
For the developer side of persistent agents, ByteForward also covered Pi 1.0 and the experimental Pi Durable framework.
Original AI generated illustration created for ByteForward. The image represents workstreams sharing a common foundation.
