Back to portfolio

nikhil-sprintly — Personal AI Productivity Agent

An AI-powered productivity and knowledge management system that manages tasks, sprints, learnings, AI tools, links, morning digests, and evening reminders through a natural-language Telegram assistant.

Agentic AIMCPMicrosoft AI Foundryn8nTelegramSupabaseLovableWeb Search Grounding

sprintly · Telegram

Agentic assistant · online

Add task: prepare steering deck, due tomorrow
✅ Added "Prepare steering deck" to Sprint 14, due tomorrow.
What's overdue?
You have 2 overdue: RAID review, vendor follow-up. Want me to reschedule them?
Message your productivity agent…
Project Overview

nikhil-sprintly is a personal AI-powered productivity and knowledge management system designed to manage tasks, sprints, learning notes, AI tools, saved links, dashboard summaries, morning digests, and evening reminders through Telegram.

Problem Statement

Friction across disconnected tools

Traditional task management requires manual updates across many tools. The project solves:

Lack of unified system for tasks, sprints, notes, learnings, and AI tools

Manual effort in tracking overdue and due-today tasks

No conversational productivity interface

No automated morning or evening planning assistant

Limited visibility into personal project progress

Solution Summary

Three engineered layers

01

Frontend & Backend Portal

Built with Lovable and Supabase to manage epics, stories, tasks, sprints, notes, learnings, AI tools, saved links, dashboard data, and user settings.

02

MCP Tool Layer

n8n exposes portal APIs as MCP tools so agents consume a standardized interface instead of calling individual APIs directly.

03

Agent & Automation Layer

Telegram assistant enables real-time interaction, while Microsoft AI Foundry agents generate morning digests and evening reminders.

High-Level Architecture

How requests flow

Interactive (Telegram)

User01
Telegram Bot02
n8n Telegram Agent Workflow03
AI Agent / MCP Client04
n8n MCP Server05
Lovable Public APIs06
Supabase Backend07

Scheduled Agents

n8n Schedule Trigger01
Microsoft AI Foundry Agent02
n8n MCP Server Tools03
Lovable APIs04
Supabase05
Foundry Agent Response06
Telegram Delivery07
Core Capabilities

What the agent can do

Add tasks through natural language

Retrieve tasks due today

List overdue tasks

Mark tasks as completed

Update story status

View current sprint progress

Save AI tools

Save notes

Save learnings

Save links

Generate dashboard summaries

Send morning digest

Send evening reminders

MCP Tools Implemented

Standardized agent tool interface

mcp_tools.ts
task_add
task_mark_done
task_get_due_today
task_get_overdue
sprint_get_current
story_update_status
dashboard_get_summary
knowledge_add_ai_tool
knowledge_add_note
knowledge_add_learning
knowledge_add_link
morning_digest_get_data
morning_digest_log_delivery
Morning Digest Automation

Automated daily planning

n8n Schedule Trigger01
Set User Context02
Get Azure Access Token03
Create Foundry Conversation04
Invoke Morning Digest Agent05
Agent calls MCP tools06
morning_digest_get_datatask_get_overduetask_get_due_today
Agent performs web search for topics of interest07
Telegram-ready digest08
n8n sends Telegram message09

Morning Digest includes

  • Good morning message
  • Overdue tasks
  • Tasks due today
  • Latest news from topics of interest
  • Suggested focus for the day
Evening Reminder Automation

Automated evening review

n8n Schedule Trigger at 7 PM01
Set User Context02
Get Azure Access Token03
Create Foundry Conversation04
Invoke Evening Reminder Agent05
Agent calls overdue and due-today task tools06
Telegram reminder message07
Key AI Concepts Used

The applied AI stack

Agentic AI

AI agents reason over requests, select tools, call APIs, and return structured responses.

MCP-Based Tool Calling

MCP standardizes backend capability access and improves tool reuse, maintainability, scalability, and separation of concerns.

AI Orchestration

n8n coordinates Telegram, Microsoft AI Foundry, MCP tools, Lovable APIs, and Supabase.

Prompt Engineering

Prompts control routing, morning digest generation, evening reminders, tool usage, safety, and output formatting.

Web Search Grounding

Morning digest uses web search for current topic updates.

Human-in-the-Loop Design

Ambiguous state-changing actions ask clarification before execution.

API-First AI Integration

Backend APIs are designed for agent consumption.

Security & Governance

Secure by design

API key validation using x-api-key

Telegram user mapping before data access

No direct database access from AI agents

MCP tools call secured backend APIs only

Sensitive details not exposed

Agent tool usage logged

Business & Productivity Impact

Measurable outcomes

Reduced manual effort in task tracking

Improved visibility of overdue and due-today work

Created daily planning habit through morning digest

Created evening review habit through reminders

Centralized learning, links, notes, and AI tools

Demonstrated practical MCP and Agentic AI implementation

Proved how agents can interact with structured systems through secure APIs

See how Agentic AI can act as a real execution assistant