Posts
All articles, thoughts, and notes.
-
Why AI Agents Need Real Email Addresses
// · 4 min readThe thesis behind AgenticMail. Why agents need their own email identity, not just API access, and how a self hosted mail server makes it possible.
-
The Outbound Guard: Preventing AI Agents from Leaking Sensitive Data
// · 5 min readHow AgenticMail's outbound guard module uses 30+ detection rules across 5 categories to stop AI agents from accidentally leaking sensitive information through email.
-
Building a Spam Filter That Protects AI Agents from Prompt Injection
// · 5 min readHow AgenticMail's spam filter uses 40+ rules across 9 categories to protect AI agents from prompt injection, social engineering, and data exfiltration attempts.
-
Defending Against Invisible Unicode Attacks in Agent Email
// · 4 min readHow AgenticMail's sanitizer strips invisible Unicode, hidden HTML elements, and other steganographic techniques that attackers use to smuggle instructions past AI agents.
-
Inter Agent Communication: Task Queues and RPC That Actually Work
// · 5 min readHow AgenticMail implements task assignment, RPC with promise based completion, and capability based access control for multi agent coordination.
-
Two Ways to Connect AI Agents to the Internet Email System
// · 4 min readAgenticMail's gateway supports relay mode via Gmail/Outlook sub addressing and domain mode with full DNS, SPF, DKIM, and DMARC configuration through Cloudflare.
-
Building an MCP Server with 62 Tools for Agent Email
// · 5 min readHow AgenticMail's MCP package exposes 62 tools over stdio transport, with stateless design, rich input schemas, telemetry, and a follow up reminder system.
-
Teaching AI Agents to Read SMS and Verification Codes
// · 5 min readHow AgenticMail's SmsManager integrates with Google Voice to give AI agents the ability to receive SMS messages and automatically extract verification codes.
-
Escalating Reminders Done Right: The Pending Email Follow Up System
// · 5 min readHow AgenticMail handles blocked outbound emails with an escalating reminder schedule, heartbeat polling, and clean process lifecycle management.
-
Sub Agent Lifecycle Management and Garbage Collection
// · 5 min readHow the OpenClaw plugin auto provisions temporary agent accounts and garbage collects stale ones on a 15 minute interval.
-
Full Text Search in SQLite: Using FTS5 for Agent Email
// · 5 min readHow AgenticMail uses SQLite's FTS5 virtual tables for fast email search, and why query sanitization matters more than you'd think.
-
Constant Time Authentication and Why It Matters
// · 5 min readHow AgenticMail uses SHA-256 hashing and timingSafeEqual for API key verification, plus a two tier key system for admin and scoped agent access.
-
The Conversational Fallacy: Why Text Based Agent Orchestration is Broken
// · 4 min readA look at the LaTeX research paper from the AgenticMail repo on structured RPC versus conversational sub agent spawning, and the 53x speedup that proves the point.
-
Building an AI Agent Workforce Platform
// · 4 min readAn overview of AgenticMail Enterprise: 6 tier architecture, 82 engine modules, 270+ tools, and the vision of treating AI agents as digital employees.
-
Agent Lifecycle State Machines: 12 States of a Digital Employee
// · 5 min readHow AgenticMail Enterprise manages agent lifecycles through a 12 state machine with health checks, auto recovery, and budget enforcement.
-
Smart Tool Loading: How We Cut LLM Token Costs by 75%
// · 5 min readAgenticMail Enterprise's 3 tier tool system that reduces baseline token usage from 33K to ~3K by loading tools intelligently based on context.
-
Data Loss Prevention for AI Agents: 53 Rules Across 7 Categories
// · 5 min readBuilding a DLP engine for Enterprise that scans agent communications in real time with 53 pre-built rules across PII, credentials, financial data, HIPAA, GDPR, intellectual property, and agent safety.
-
Building Compliance Reports for AI Agent Operations
// · 5 min readFive report types for AI agent compliance: SOC 2 Type II, GDPR DSAR, SOX Audit Trail, Incident Report, and Access Review. Full HTML export with risk scoring and control effectiveness analysis.
-
The Action Journal: Every Agent Decision with Full Rollback
// · 4 min readHow AgenticMail Enterprise journals every agent action with before/after state snapshots, actor identity, and timestamps, plus full rollback capability for when things go wrong.
-
Knowledge Base and RAG for Enterprise AI Agents
// · 5 min readDocument ingestion, chunking, BM25F search, automatic context injection, and per-agent access control. How RAG works differently when agents are the consumers.
-
Workforce Management for AI: Shifts, Rotations, and Capacity Planning
// · 5 min readTreating AI agents like employees with shift schedules, on-call rotations, capacity planning, clock records, off-duty enforcement, vacation auto-responders, and birthday automation.
-
Building a Prediction Market Trading System Inside an Agent Platform
// · 5 min read126 tools across 10 skill modules, three trading modes, quantitative analysis with Kelly criterion and Black-Scholes, on-chain intelligence, and risk management with circuit breakers.
-
Teaching AI Agents to Join Google Meet and Speak
// · 5 min readMeeting intelligence through Playwright browser automation, ElevenLabs TTS through virtual audio devices with sox, real-time transcription, and meeting notes generation.
-
145 SaaS Integrations: Building a Universal Tool Adapter Pattern
// · 5 min readThe adapter pattern behind 145 SaaS integrations. Each adapter has tool definitions, API executor, credential resolution from Vault, OAuth flow, and rate limit handling.
-
Soul Templates: Giving AI Agents Personality and Purpose
// · 4 min readHow 51 soul templates across 14 categories transform generic AI into specialized agents with real personality and domain expertise.
-
Building a React Dashboard Without a Build Step
// · 4 min read28 pages and 23 agent detail tabs served as a single HTML file. No webpack, no vite, no bundler. Just React 18 from a CDN.
-
One Schema, Ten Backends: Cross Database Migration
// · 5 min readSupporting 10 database backends behind one adapter interface, with automatic DDL conversion and smart environment detection.
-
Agent Autonomy: Morning Triage, Daily Catchup, and Goal Tracking
// · 5 min readHow AgenticMail Enterprise agents manage their own schedules, triage inboxes, track goals, and operate independently with heartbeat monitoring.
-
Secure Vault: Encrypted Credential Storage for AI Agents
// · 5 min readEncrypted credential storage with org scoped namespacing and automatic resolution through the tool adapter pattern.
-
Agent to Agent Communication: Messages, Broadcasts, and Channels
// · 5 min readHow AI agents in AgenticMail Enterprise coordinate through direct messages, broadcasts, topic channels, and priority levels.
-
Domain Lock: Cryptographic Deployment Protection
// · 4 min read256 bit deployment keys, DNS TXT verification, and offline operation. How AgenticMail Enterprise prevents unauthorized deployment duplication.
-
The Nine Layer Middleware Stack: Securing an Agent API
// · 5 min readA Hono API server with 9 middleware layers: request ID, transport encryption, security headers, CORS, rate limiting, IP firewall, audit logging, RBAC, and org scoping.
-
Designing a REST API with 107 Endpoints for AI Agent Email
// · 5 min readHow the @agenticmail/api package structures 107 endpoints across 11 route modules using Express, factory patterns, and connection pooling.
-
Building a Developer Friendly CLI with an Interactive Shell
// · 5 min readHow the agenticmail CLI balances developer experience with system management through animated spinners, setup wizards, and an interactive shell.
-
Building an OpenClaw Plugin: Hooks, Channels, and Auto Provisioning
// · 5 min readHow the @agenticmail/openclaw package integrates AgenticMail into OpenClaw with 52 tools, lifecycle hooks, and automatic sub agent provisioning.
-
Structuring a TypeScript Monorepo: Five Packages, One Product
// · 5 min readHow AgenticMail's monorepo uses npm workspaces, tsup, and a clean dependency graph to ship five packages from one repository.
-
Defense in Depth: Client Side and Server Side Outbound Scanning
// · 5 min readWhy AgenticMail scans outbound email at two layers: 35 regex patterns in the MCP client and a blocking guard in the API server.
-
Real Time Email Notifications with Server Sent Events
// · 5 min readHow AgenticMail uses SSE and IMAP IDLE to push new email notifications to agents in real time, with inline spam filtering and rule evaluation.
-
Email Templates with Variable Substitution and Natural Language Scheduling
// · 5 min readHow AgenticMail's features route handles templates with {{var}} substitution, natural language scheduled sends, drafts, signatures, and tag based organization.
-
Giving AI Agents Their Own Database: A Full DBMS Over REST
// · 6 min readHow AgenticMail's storage route exposes 28 database operations over REST, giving AI agents DDL, DML, index management, and import/export capabilities.
-
From Database Administration to AI Agent Coordination
// · 4 min readHow years of optimizing MySQL, PostgreSQL, and IBM DB2 platforms shaped the way I build AI agent systems today.
-
Using Data to Optimize Waste Collection and Recycling Operations
// · 4 min readHow I helped a solid waste management company use data analytics to improve collection routes, scheduling, and recycling rates.
-
What Years of Excel Taught Me About Working with Data
// · 5 min readThe underrated power of spreadsheets, when they're the right tool, and how financial analysis in Excel made me a better engineer.
-
Building an End to End Machine Learning Pipeline for Airfoil Noise Prediction
// · 5 min readA walkthrough of the PySpark ML pipeline I built using the NASA Airfoil Self Noise dataset, from ETL to model evaluation.
-
Why I Volunteer: From COVID 19 Campaigns to Campus Community Building
// · 4 min readReflections on volunteering with the H2H vaccination campaign and Carolina University campus initiatives, and why giving back matters in tech.