CoAgent4U Logo
CoAgent4U

A privacy-centric personal agent that manages your calendar, tasks, and productivity through MCP tools โ€” and coordinates with other agents via A2A protocol.

MCP ProtocolA2A ProtocolLLM-Driven Planning32 MCP Tools
CoAgent4U
Connected
TZ
Tanmay Zade4:24 PM

CoAgent4U
CoAgent4UAPP4:24 PM

Coordinating with Sarah's agent. Common availability found: 2:00 PM โ€“ 5:00 PM. Awaiting Sarah's approval before confirming.

Agent-to-Agent coordination in progress
CoAgent4U
CoAgent4UAPP4:24 PM
๐Ÿ“…

Proposed Slots

I've proposed these available time slots to @Sarah. Waiting for their selection...

๐Ÿ—“๏ธFri, Mar 21, 2026
CoAgent4U
CoAgent4UAPP4:25 PM
๐Ÿ“‹

Meeting Approval Request

@Sarah selected a meeting slot.

Fri, 14 Mar 2026
๐Ÿ•03:00 PM โ€“ 04:00 PM

Approve or reject this meeting time.

CoAgent4U
CoAgent4UAPP4:26 PM
โœ“

Meeting Confirmed

Participants:

  • @Tanmay Zade
  • @Sarah
Fri, 14 Mar 2026
๐Ÿ•03:00 PM โ€“ 04:00 PM

The Architecture

One Agent. Two Protocols. Infinite Capability.

CoAgent4U gives you a personal AI agent that accesses your tools via MCP and collaborates with other agents via A2A โ€” all with you in control.

MCP โ€” Tool Access Layer

32 tools across 3 MCP servers โ€” Calendar (17), Tasks (8), and Productivity (7). Your agent accesses Google Calendar & Tasks through the standard Model Context Protocol.

A2A โ€” Agent Collaboration

Agents negotiate scheduling across users using the Agent-to-Agent protocol. Your agent talks to others' agents to find common availability and coordinate meetings.

Privacy-First Design

Human-in-the-loop approvals for every action. Your agent proposes, you decide. Data stays yours โ€” encrypted at rest, with PII detection and data minimization.

32 MCP Tools

Your agent, your tools, your rules

Your personal agent connects to 3 MCP servers with 32 tools for calendar, tasks, and productivity โ€” all accessible through natural language.

Calendar via MCP

17 MCP tools for Google Calendar โ€” list events, check availability, create/delete events, find free slots, all through the standard MCP protocol.

Tasks via MCP

8 MCP tools for Google Tasks โ€” create task lists, add/update/complete/delete tasks. Your agent manages your to-do list through natural language.

LLM-Driven Planning

Groq-powered LLM planner decides which MCP tools to call based on your message. It plans multi-step tasks and summarizes results in natural language.

A2A Coordination

Your agent communicates with other users' agents via the Agent-to-Agent protocol to find common availability and schedule meetings across calendars.

Productivity Intelligence

7 MCP productivity tools โ€” daily briefings, weekly summaries, focus time blocks, smart scheduling with preferences, and recurring routines.

Conflict Detection

Automatic conflict detection across all your commitments before creating events. Your agent checks availability and prevents double-bookings.

MCP + A2A Architecture

How MCP tools and A2A coordination work together

Your personal agent uses MCP to access your tools and A2A to collaborate with other agents โ€” all orchestrated by an LLM planner.

User Message โ†’ LLM Planner โ†’ MCP Tool Calls โ†’ A2A Negotiation โ†’ Human Approval โ†’ Execution

1

User Message

User sends a natural language request via Slack or direct HTTP (e.g. "schedule a meeting with Sarah").

2

LLM Plans Tool Calls

Groq LLM receives 32 MCP tool schemas and decides which tools to call based on user intent.

3

MCP Tool Execution

Agent executes planned tools via MCP clients โ€” reading calendar, checking availability, creating events.

6

Invitee Selection

Invitee's agent presents available slots via Slack. The invitee selects a preferred time.

5
A2A

A2A Slot Proposal

Agent A sends proposed slots to Agent B via A2A protocol. Both agents' availability is compared.

4
A2A

A2A Initiation

Agent initiates A2A session with invitee's agent to negotiate meeting times across both calendars.

7

Human Approval

Requester reviews the selected slot and explicitly approves or rejects the meeting via Slack.

8

MCP Calendar Create

Agent calls MCP calendar tools to create the event on both Google Calendars via OAuth.

9

Confirmed & Synced

Meeting confirmed. Both agents sync calendars via MCP. LLM summarizes outcome to both users.

Real Use Case Examples

See it in action

Simple commands, powerful coordination. Walk through complete flows to see how agents work.

# coagent4u-demo
Connected
TZ
Tanmay Zade11:33 AM

show my schedule

CoAgent4U
CoAgent4UAPP11:33 AM
๐Ÿ“…

Your upcoming schedule (next 7 days):

  • โ€ข Team StandupMon, Mar 20, 09:00 am โ€“ 09:30 am
  • โ€ข Design ReviewMon, Mar 20, 10:30 am โ€“ 11:30 am
  • โ€ข Lunch with AlexMon, Mar 20, 01:00 pm โ€“ 02:00 pm
  • โ€ข Sprint PlanningMon, Mar 20, 03:00 pm โ€“ 05:00 pm
  • โ€ข Team LunchTue, Mar 21, 12:00 pm โ€“ 01:00 pm

Why MCP + A2A

Built on open protocols, not proprietary locks

CoAgent4U combines MCP for tool access and A2A for agent collaboration โ€” giving you extensibility without vendor lock-in.

MCP-Based Extensibility

Adding a new tool is just adding another MCP server. No agent code changes. Calendar, Tasks, Drive, Gmail โ€” any resource becomes an MCP tool.

Agent-to-Agent Collaboration

Your agent talks to others' agents via A2A. Cross-user scheduling happens automatically without manual back-and-forth communication.

LLM-Orchestrated Planning

The Groq-powered LLM planner sees all 32 MCP tool schemas and decides the optimal execution plan for each user message.

Hybrid Architecture

Java Spring Boot backend for A2A coordination + Python FastAPI agent for MCP tools and LLM planning. Best of both worlds.

Human-in-the-Loop Control

Every coordination action requires your explicit approval via Slack. Your agent proposes, you decide. Nothing happens without consent.

Agent Security Status

All systems secure

OAuth2 IntegrationConnected
MCP Tool Servers3 Active (32 tools)
A2A ProtocolInitialized
Data EncryptionAES-256
Agent Activity LoggingEnabled

LLM Model: Groq Llama 3.3 70B ยท Agent Runtime: FastAPI + FastMCP

Security & Trust

Privacy-centric by design

Your personal agent handles sensitive data โ€” calendar events, task lists, meeting details. Every layer is built with security first.

OAuth2 per Agent

Each agent's Google access is scoped via OAuth2. Tokens stored encrypted in PostgreSQL via the Java bridge, never in files.

MCP Tool Isolation

MCP servers run in-process with separate tool clients. Each tool call is scoped to the authenticated agent โ€” no cross-agent data leaks.

AES-256 Encryption

All agent memory and OAuth tokens encrypted at rest using AES-256. Conversation context is encrypted before writing to the database.

Full Activity Logging

Every tool call, A2A interaction, and LLM plan is logged via the Java bridge. Full transparency into what your agent does and when.

Get Started Today

Your Personal Agent, Powered by MCP & A2A

Stop managing calendars manually. Let your personal agent access tools via MCP, coordinate with others via A2A, and keep you in control.

Test users onlyeasychat148@gmail.com