Thred

What it does

i built Thred because AI agents keep forgetting what happened before them. Thred is a shared memory and coordination layer for AI agents. It gives agents one workspace for task context, decisions, progress, evidence, and handoffs instead of keeping everything trapped inside separate chat sessions. A user gives one task to Thred’s Agent Studio. Thred creates a working team where product planning, frontend work, backend work, and QA review happen in parallel. Each agent leaves a structured handoff in the shared workspace, so the other agents can continue from the same context. The result is a coordinated team workflow instead of four agents producing disconnected answers.

How the agents run concurrently

The user submits one task through Thred’s Agent Studio. Thred creates one Mozaik runtime with four specialized agents: 01 / Product Lead Breaks the task into requirements, acceptance criteria, and implementation decisions. 02 / Frontend Builder Plans the interface, user flow, and frontend changes required for the task. 03 / Backend Builder Plans the API, data model, server-side logic, and integration work. 04 / QA Reviewer Reviews the shared work, identifies risks, checks whether the requirements are covered, and makes the final release recommendation. All four agents start their Mozaik loops concurrently. They share the same Thred workspace state, including the original task, agent events, handoffs, artifacts, findings, and QA release status. The agents do not simply return four isolated chatbot messages. Each agent writes a structured handoff that becomes available to the rest of the team. The QA agent reviews the combined output and produces a release decision based on the shared work. Thred stores the run history and handoffs in PostgreSQL so the work can be reopened instead of disappearing when the session ends. Mozaik Cloud provides live visibility into the running agent loops and their events.