AI Task Planning for ADHD
We built a memory-aware AI companion for ADHD task planning. It turns scattered thoughts and changing priorities into actionable plans. LangGraph and auto-compacting memory retain context without slowing repeated interactions.
Comigo needed useful continuity across changing conversations. The agent had to retain long-term memory, recognize recurring patterns, reprioritize tasks, and control token cost as history grew. Responses had to stay immediate on the web and in a Chrome extension.
AI Task Planning for ADHD
Comigo combines a web app, AI orchestration layer, and browser companion. FastAPI handles sessions, task workflows, and AI services. SvelteKit keeps the interface responsive. The Chrome extension works while users browse. LangGraph drives streaming, tool calling, and multi-step reasoning over OpenAI models. Neon stores users, chat logs, tasks, and memory summaries. Stripe handles payments and subscriptions.
SvelteKit keeps the frontend and Chrome extension light. LangGraph adds streaming and tool calling to controlled execution paths. Conversation history compacts into milestones before each request. That preserves context while reducing tokens and latency. Docker runs on GCP, and Neon autoscaling removes database provisioning work.
Streaming responses make planning feel immediate. LangGraph tool calls run structured workflows. Persisted chat history and compacted summaries retain context. Tasks reprioritize as user input changes, and planning flows break work into smaller steps. The Chrome extension stays available during browsing. Stripe handles subscriptions without storing card data on Comigo servers.
Launched the MVP on Product Hunt and AppSumo in about two months.
Delivered ADHD task support with persistent, memory-aware chat.
Validated task planning that reprioritizes with user context.
Shipped streaming responses and tool-calling workflows.
Made the assistant available on web and Chrome extension with token-efficient memory.
An ADHD assistant needs context, continuity, and changing priorities. A static task manager cannot provide that support.
Memory must be compact, relevant, and inexpensive to send back to the model.
Agent output depends on when, how, and what context reaches the model.
Memory compaction controls long-context token cost and latency.
A focused SvelteKit and LangGraph stack can serve web and extension surfaces consistently.
Relevant to AI companions, productivity agents, mental-wellness tools, coaching platforms, and workflow assistants. It fits products that need persistent memory and adaptive task execution.
We specialize in ai integration and mvp for productivity / wellness companies. If you're facing challenges like the ones we solved for Comigo, let's talk.
