Features
39 capabilities powered by Claude Agent SDK
WhatsApp Integration
Receives and sends messages via WhatsApp Cloud API. Supports text, media, audio, documents and templates. The main communication channel with Djalma.
Memory System
Long-term memory with smart search. Stores personal data, preferences, documents and recurring information. Never forgets what matters.
Calendar & Reminders
Creates, lists and manages events and reminders in Google Calendar. Direct CLI integration for maximum speed.
Email (Gmail)
Reads, searches and drafts emails via Gmail API. Monitors the inbox and automatically notifies about important messages.
Google Drive
Lists, searches, reads and downloads files from Google Drive. Quick access to spreadsheets, documents and any stored file.
TDAH Focus Manager
Daily planning with the Rule of 3: 1 Must-Do + 2 Important tasks. Task inbox and end-of-day protocol designed for ADHD.
Payments & Banking
Tracks bills, due dates and balances for Nubank and Bradesco. Full financial history at your fingertips.
Loans (Poupançudo)
Manages personal loans with monthly compound interest. Tracks debtors, contributions and updated balances.
Trips
Complete travel management: flights, accommodation and automatic reminders. Sends check-in alerts and packing lists at the right time.
Projects Kanban
Kanban boards with planner, coder, debugger and reviewer agents. Plans and implements software tasks autonomously.
Scheduler
Creates scheduled tasks via cron that run periodically. Results are delivered directly to WhatsApp with automatic notifications.
Daily Summary
Automatic morning briefing with schedule, emails, weather and daily tasks. Start every day informed and organized.
Token Monitor
Monitors token usage and cost per query, hour and day. Sends WhatsApp alerts when consumption exceeds configured limits.
Subprocess Agent
Delegates long or heavy tasks to isolated Claude subprocesses. Frees the main context and notifies when the work is done.
History
Searches and navigates previous conversation history. Quickly finds what was said, decided or planned in past sessions.
Speedtest
Tests internet speed on demand. Returns ping, download and upload in seconds.
Lessons Learned
Automatically captures user corrections and learnings. Reuses these insights in future tasks to avoid repeating mistakes.
Media Handler
Receives and processes images, photos and media sent via WhatsApp. Supports QR codes, receipts and screenshots.
WhatsApp Templates
Manages WhatsApp Business API templates. Creates, lists and sends standardized messages with Meta approval.
Session Monitor
Displays current session info: tokens used, estimated cost and compactions. Full transparency over conversation consumption.
Version & Updater
Didi version control with detailed changelog. Updates via git and records all changes made.
Browser Automation
Headless browser automation with Playwright. Navigates websites, takes screenshots, fills forms and extracts data from any web page.
Claude Sessions
Lists and monitors active Claude Code sessions running on the server. Shows token usage, session activity and cost per session.
GOWA (Personal WhatsApp)
Sends and receives messages from Didi's own personal WhatsApp number via GOWA integration — enabling two-way personal messaging beyond the Business API.
Audio Narrator
Converts articles, URLs or any text into narrated audio. Generate podcast-style recordings and listen to content on the go via text-to-speech.
AI Image Generation
Generate or edit images with AI. Create illustrations, photos, art and designs from natural language descriptions sent via WhatsApp.
Redraw Tutorial
Send any image and get a step-by-step drawing tutorial. Breaks down complex images into simple, followable drawing instructions.
File Manager
Browse and retrieve generated images, drawings and tutorials. Search through creation history and view prompts used for past generations.
Balance Tracker
Manually record and query bank balances across multiple accounts. Get a financial snapshot comparing available balance vs pending expenses.
Video Info
Fetch metadata and details from YouTube, Vimeo or any video URL. Get titles, descriptions, duration and channel info without leaving WhatsApp.
Self-Scheduler
Didi schedules future tasks for itself with full context preserved in a mailbox. Enables autonomous retries, follow-ups and time-dependent workflows.
Income Tracker
Track salary, freelance and multi-currency income in BRL, USD, EUR and GBP. Monthly balance comparing income vs expenses.
Wise Integration
Parse Wise PDF statements and import transactions across BRL, USD, EUR and GBP accounts. Integrated with income and payment tracking.
Transaction Reconciliation
Interactive WhatsApp flow to categorize unmatched bank transactions. AI suggests categories automatically — only asks when uncertain.
Dev Docs
Internal architecture reference for Didi's codebase — file structure, CLI commands and patterns consulted before implementing changes.
Interactive Menu
WhatsApp interactive button menu with usage tracking. Frequent features surface automatically as favorites for quick access.
Git Save
Commit and push all pending codebase changes via background agent. Keeps Didi's codebase in sync with a single natural language command.
Automations
Works in the background even when you're not talking to it
Architecture
How DiDi works under the hood
Message arrives via WhatsApp
Meta Cloud API via n8n webhook forwards to the FastAPI server over Tailscale VPN.
Claude Agent SDK processes it
Persistent streaming session. System prompt loads essential tools; skills are activated on demand via prompt caching.
Tools and MCPs execute
Google Calendar, Gmail, payments database, memory manager, trip system — via Python tools and MCP servers.
Response formatted for WhatsApp
Formatted for quick reading on mobile. Supports images, audio, documents, locations, and interactive messages.
What makes it different
Self-Healing
Auto-restart via supervisord, auto-update via Git, and AI-powered conflict resolution.
Persistent Memory
Remembers personal information, compacts history daily, and learns your preferences over time.
Proactive
Doesn't wait to be asked — sends alerts, summaries, and check-ins automatically.
Private
Runs on your own server via Tailscale VPN — your data never passes through third parties.