ASI:One

What ASI:One combines

ASI:One is a personal AI workspace from Fetch.ai that combines conversation with access to specialist agents. Its consumer interface lets you bring specialized AIs into a conversation through mentions and coordinate related work in one place. The service also has a developer API for applications that need model responses, tool calls, or multi-step planning.

It is relevant to people organizing ongoing personal projects and to developers exploring agent-backed applications. These are different ways to use the product: a person can work through the chat interface, while a developer integrates model and planning capabilities into software. Access to an agent network does not mean every specialist can perform every requested action.

Projects, recurring tasks, and specialist help

Projects group the chats and recurring tasks belonging to one effort, such as a job search, trip, or side project. Chats can be created inside a project or moved into it; linked recurring tasks come with their conversation. This organization helps keep a recurring update beside the discussion that explains why it matters.

The recurring-task feature describes background discovery of connections, activities, and opportunities based on preferences. Specialist AIs can be mentioned in chat for their particular services. Review each specialist's role and requested permissions rather than assuming all services share the same terms, costs, or reliability.

Developer capabilities and an example workflow

The developer platform offers an OpenAI-compatible API with three model choices: asi1, asi1-mini, and asi1-ultra. The official model guide distinguishes their capabilities and resource tradeoffs. Developers can request structured data, stream responses, or define their own tools instead of relying solely on free-form text.

Planner mode breaks a request into steps and can discover independent Agentverse agents. For example, a developer building an event-research assistant could request a shortlist for specified dates and location, stream the plan's progress, and present candidate results for review. Check dates and source details before taking any booking action. If the planner returns a custom tool call, the application must execute that tool and return its result; the model does not automatically run your local function.

Limits, access, and costs

Independent agents can be slow, decline a request, or fail to reply. Planner runs can pause for a clarification, payment request, or confirmation, and have bounded planning rounds. An API application must preserve the documented session identifier when resuming a paused plan. A final answer can contain only the work completed before the limit, so completion of the response is not proof that every requested step succeeded.

Check consumer account terms on the official site before relying on ongoing access or a particular specialist. The developer model documentation publishes usage-based token pricing; model use and any paid agent service should be considered separately. Compare the actual work returned with your request, review proposed payments, and verify important generated information. Personalization and project context can make a conversation more relevant without making its conclusions automatically correct.