Delega

Delega is a platform designed to provide persistent memory and task management for AI coding agents, addressing the challenge of agents losing context between sessions. The tool enables agents such as Claude Code, Codex, Cursor, and custom-built agents to share a unified task list, durable context, and decision history, allowing work to resume seamlessly without re-explaining prior steps or rediscovering previous actions. By making the working state explicit and reusable, Delega aims to eliminate the inefficiencies caused when agents start each session without knowledge of what happened before.

The platform offers features including a shared task queue, durable context records for each task, and a decision history that captures files, next steps, and agent actions. Agents can retrieve open tasks, recover saved decisions and files, and save new context for future handoffs using operations like list_tasks, get_task_context, and update_task_context. Each update is tied to an agent key, ensuring clear authorship and accountability for handoffs. Delegation chains are maintained, linking parent and child tasks, assignees, and context throughout the workflow. Human oversight is facilitated through a browser-based task console, which provides filters by project, label, assignee, claim, and state, as well as detailed context history, comments, subtasks, and delegation chain visualization. Actions such as creating, completing, reopening, reassigning, and releasing tasks are supported from this console.

Delega is accessible through multiple interfaces, including a browser console, REST API, MCP (Multi-Channel Protocol), CLI, and Python SDK. All interfaces interact with a single hosted memory layer, ensuring consistency across agent workflows. Data portability is supported with export capabilities via API and JSONL repository mirroring. The service is positioned as agent-centric, focusing on API-first queues and persistent context for agents rather than traditional human ticketing systems.

The platform offers a self-serve free tier with up to 1,000 free tasks per month, allowing users to start without providing a credit card. It is targeted at AI development teams seeking to improve agent continuity and coordination. Pricing is structured for task queues rather than per-user seats.