AI Project Showcase: My TravelPlanner

Document type: AI Project Showcase

Project: My TravelPlanner

Status: Draft

Last updated by Claude Code: April 12, 2026

Populated from: CLAUDE.md (root), .claude/CLAUDE.md, my-travelplanner/my-travelplanner.php, my-travelplanner/composer.json, MyTravelPlanner/project.yml, documentation/N8N-DIFY-GUIDE.md, requirements/skills-and-agents.md, MyTravelPlanner/MyTravelPlanner/Services/SystemPromptBuilder.swift, .claude/settings.json, tests/pytest.ini, tests/requirements.txt, browser-extension/manifest.json, tests/smoke-test-plan.md

Section 1 — Product Overview

1.1 Product name and tagline

Name: My TravelPlanner Tagline: AI-powered travel planning platform — plan personalized trips with intelligent recommendations Current status: Active development — WordPress plugin + iOS/macOS native app in parallel (Build 23, March 2026) First commit / project start: March 2026 (initial monorepo commit 3a82f49a on March 10, 2026)

1.2 What it is

My TravelPlanner is a multi-platform AI-powered travel planning product that helps leisure travelers plan personalized trips through conversational AI. It combines a WordPress plugin with an AI chatbot, a native iOS/macOS companion app built in SwiftUI, and a Chrome browser extension for saving places from travel sites. The system uses 17 specialist AI advisor agents for travel, scuba, expat, RV, culinary, hotel, weather, airport, transit, rental, navigation, remote work, retirement, civil unrest, and FIFA 2026 World Cup planning. The platform is backed by a 193-file knowledgebase covering destinations, activities, and specialized travel domains, with n8n workflow orchestration and Dify-managed vector retrieval via pgvector.

1.3 What makes it meaningfully different

  • 17 specialist advisor agents with full PHP/Swift parity across platforms — not a single-purpose chatbot but an orchestrated multi-agent system
  • Three-layer AI pipeline: n8n workflow orchestration first, with automatic fallback to local specialist agents, with Claude API as the final backend
  • Cross-platform parity: WordPress web, iOS, and macOS share the same knowledgebase and agent architecture
  • P0–P6 budget-controlled prompt assembly (100K char / ~25K token budget) with prioritized knowledge injection
  • Browser extension for saving places directly from Google Maps, TripAdvisor, Yelp, Booking.com, and Airbnb

1.4 Platform and deployment context

Platform: WordPress plugin (web) + native iOS/macOS app (SwiftUI) + Chrome browser extension (MV3) Deployment: WordPress self-hosted, iOS App Store, Mac App Store (planned) Primary interface: Conversational chatbot (web widget + native app chat), Kanban itinerary boards, route planning maps


Section 2 — User Needs and Problem Statement

2.1 Target user

Primary user: Leisure travelers seeking intelligent trip planning assistance Secondary users: Expats/relocators, RV travelers, scuba divers, digital nomads, FIFA 2026 attendees, retirees considering relocation User environment: Web browsers (WordPress site), iOS devices, macOS desktop

2.2 The problem being solved

Trip planning involves coordinating multiple complex, interrelated decisions (destinations, lodging, dining, activities, budgets, weather, safety) across fragmented information sources. Travelers currently piece together plans from dozens of websites, review platforms, and travel guides with no unified intelligence layer. My TravelPlanner consolidates this into a single AI-powered advisor that understands the user’s preferences, constraints, and context.

2.3 Unmet needs this addresses

Need How the product addresses it Source of evidence
Personalized multi-domain travel advice 17 specialist agents for different travel domains CLAUDE.md agent inventory
Real-time information alongside curated knowledge Tavily web search + 193-file KB + Pinecone vector search SystemPromptBuilder.swift, N8N-DIFY-GUIDE.md
Cross-device trip access CloudKit sync between iOS/macOS + WordPress web access .claude/CLAUDE.md feature list
Event-specific planning Dedicated FIFA 2026 advisor with 25-file city-by-city knowledgebase knowledgebase/fifa-2026/
Save-from-anywhere workflow Browser extension captures places from Google Maps, TripAdvisor, Yelp, Booking, Airbnb browser-extension/manifest.json

2.4 What users were doing before this existed

Using multiple separate tools — Google Maps, TripAdvisor, Booking.com, spreadsheets, travel blogs — with no AI-powered synthesis. Planning required manual research across siloed platforms.


Section 3 — Market Context and Competitive Landscape

3.1 Market category

Primary category: AI-powered travel planning platforms Market maturity: Emerging — many AI travel tools exist but few offer multi-agent specialist depth [CLAUDE NOTE: inferred] Key dynamics: Rapid entry of LLM-powered travel assistants; differentiation through domain depth and multi-platform delivery [CLAUDE NOTE: inferred]

3.2 Competitive landscape

Product / Company Approach Strengths Key gap this project addresses Source
⚡ Google Travel Aggregation + search Scale, maps integration No conversational AI advisor, no specialist agents General knowledge
⚡ Tripit Itinerary organization Email parsing, calendar sync No AI planning, no recommendations General knowledge
⚡ Wonderplan / Roam Around AI trip generation Quick itinerary creation Shallow single-model responses, no domain specialization General knowledge

3.3 Market positioning

Multi-agent specialist travel advisor with cross-platform delivery (web + native) and deep domain knowledge (scuba, expat, RV, FIFA 2026). Positioned as a premium planning tool rather than a quick-itinerary generator.

3.4 Defensibility assessment

Defensibility comes from the 193-file knowledgebase, 17 specialist agent configurations with tuned prompts, the multi-platform architecture (WordPress + SwiftUI), and the n8n/Dify integration layer. The budget-controlled prompt assembly system (P0–P6 priority tiers) is a reusable architectural pattern. [CLAUDE NOTE: inferred from code architecture]


Section 4 — Requirements Framing

4.1 How requirements were approached

Extensive requirements documentation exists across documentation/ (detailed requirements at ~2700+ lines, UI design spec, test plan) and requirements/ (17 files covering feature parity, competitor analysis, data sources, agent/skill specifications). Requirements evolved through iterative AI-assisted development sessions.

4.2 Core requirements

  1. Conversational AI trip planning via Claude API with multi-agent routing
  2. 17 specialist advisor agents with shared knowledgebase
  3. WordPress plugin with chatbot widget and admin settings
  4. Native iOS/macOS app with SwiftUI, CloudKit sync, and widgets
  5. n8n workflow orchestration with Dify vector retrieval
  6. Kanban itinerary boards, route planning, budget tracking
  7. Browser extension for saving places from travel sites
  8. Feature parity between web and native platforms

4.3 Constraints and non-goals

Hard constraints: WordPress 6.0+ / PHP 8.0+ (plugin); iOS 17.0+ / macOS 14.0+ (native); Swift 5.9; Mapbox Maps 11.20.0 dependency Explicit non-goals: NOT FOUND — add manually

4.4 Key design decisions and their rationale

Decision Alternatives considered Rationale Evidence source
Multi-agent architecture (17 agents) Single general-purpose chatbot Domain specialization yields higher quality advice per topic CLAUDE.md agent inventory
n8n-first with local fallback Direct API calls only Centralized orchestration, easier workflow updates, graceful degradation N8N-DIFY-GUIDE.md
P0–P6 budget-controlled prompts Fixed prompt templates Dynamic assembly maximizes context within token limits SystemPromptBuilder.swift
XcodeGen for project management Manual Xcode project Reproducible builds, clean diffs, declarative config project.yml
Mapbox Maps over Apple Maps Apple Maps, Google Maps Richer customization, cross-platform potential project.yml dependency

Section 5 — Knowledge System Architecture

5.1 Knowledge system overview

KB type: Curated markdown files + Dify-managed vector database (pgvector) + Pinecone embeddings + Tavily web search Location in repo: knowledgebase/ (193 markdown files), loaded as Xcode resources via project.yml Estimated size: 193 markdown content files across 10+ categories; 26-file FIFA 2026 sub-corpus (9 top-level topic files + 11 US city + 3 Mexico city + 2 Canada city + 1 tavily-domains.md); Dify indexes 183+ files

5.2 Knowledge system structure


knowledgebase/
├── fifa-2026/              # FIFA 2026 World Cup (26 files)
│   ├── overview.md, schedule.md, venues.md, teams.md
│   ├── travel-logistics.md, tickets-budget.md, transportation.md
│   ├── food-entertainment.md, safety-health.md, tavily-domains.md
│   ├── us-cities/          # 11 US host city guides
│   ├── mexico-cities/      # 3 Mexico host city guides
│   └── canada-cities/      # 2 Canada host city guides
├── countries/              # Country and city guides
├── travel/                 # General travel topics
├── scuba/                  # Scuba diving (shared with Scuba GPT)
├── expat/                  # Expat/relocation guides
├── rv-camping/             # RV and camping (incl. international)
├── airports/               # Airport guides
├── remote-work/            # Digital nomad content
├── accommodation/          # Lodging guides
├── destinations/           # Regional destination guides
├── international/          # International travel topics
├── roadtripping/           # Road trip guides
├── travel-guides/          # Compiled travel guides
└── seasonal-planner.md     # Seasonal planning reference

5.3 Knowledge categories

Category Files / format Purpose Update frequency
FIFA 2026 26 .md files World Cup host city guides, logistics, scheduling, tavily domain filters Active updates for 2026 event
Countries & cities Multiple .md Destination profiles, cultural info, travel tips Periodic
Scuba diving Regional .md guides Dive site info, seasonal conditions Shared with Scuba GPT product
Expat/relocation .md guides Relocation logistics, visa info, cost of living Periodic
RV/camping .md guides Route planning, campground info, international RV Periodic
Airports .md guides Terminal info, transit connections, layover tips Periodic
Remote work .md guides Digital nomad destinations, coworking, visa programs Periodic
Accommodation .md guides Lodging types, booking strategies Periodic

5.4 How the knowledge system was built

Curated markdown files authored manually and through AI-assisted research sessions. FIFA 2026 corpus built as a dedicated sub-collection with city-by-city guides for all 16 host cities across US, Mexico, and Canada. Knowledge is loaded at runtime through keyword-based matching (P3 tier) and Dify semantic retrieval via pgvector embeddings.

5.5 System prompt and agent configuration

System prompt approach: P0–P6 priority-based budget assembly in SystemPromptBuilder.swift (548 lines):

  • P0: Agent system prompt (base persona + role)
  • P1: User profile context (2K budget)
  • P2: Trip context (4K budget)
  • P3: KB keyword matches (60K) + Pinecone vectors (4K)
  • P4: Tavily web search results (8K)
  • P5: Conversation history as messages (12K)
  • P6: Safety guardrails (2K)
  • Total cap: ~100K characters (~25K tokens)

Key behavioural guardrails:

  • Scuba: DAN phone number required, no deco/gas mix/medical clearance advice
  • Expat/local guide: Safety and legal disclaimers
  • Place hotlink syntax: [[place:Name, City]]
  • Itinerary block syntax: [[itinerary:{JSON}]]

Persona / tone configuration: Rich per-agent personas configured in Swift and PHP agent classes Tool use / function calling: NOT FOUND in current prompts — agents use RAG context injection rather than tool calling


Section 6 — Build Methodology

6.1 Development approach

AI-assisted parallel development: WordPress plugin and iOS/macOS app developed simultaneously with shared knowledgebase. Claude Code / Cursor used for code generation, architecture decisions, and documentation. n8n/Dify infrastructure added as an orchestration layer in March 2026.

6.2 Build phases

Phase Approximate timeframe What was built Key commits or milestones
Initial commit Mar 10, 2026 Coding playbook, Claude Platform Skills, portfolio 3a82f49a
Pre-migration snapshot Mar 27, 2026 Full workspace captured before n8n+Dify migration e65fdaf2
n8n + Dify infrastructure Mar 27, 2026 Docker infrastructure (Phase 1) fc3cecce
Agent/skill expansion Mar 28, 2026 Expanded agent roster, skills library, tests f8505ec4
RV + UI + n8n integration Mar 31, 2026 RV lifestyle features, UI design system, n8n integration 44dff0a8
Cross-product updates Mar 31, 2026 n8n workflow client and agent updates across products 82aee6a6
Smoke testing Apr 1, 2026 Production smoke test, defects D001–D005 documented tests/smoke-test-plan.md

6.3 Claude Code / AI-assisted development patterns

  • Cursor IDE for primary plugin and app development
  • Claude Code for CLAUDE.md management and cross-file refactors
  • AI-generated knowledgebase content (193 files)
  • AI-assisted requirements documentation (2700+ line detailed requirements)
  • Pytest test suite generation for API endpoint validation

6.4 Key technical challenges and how they were resolved

Challenge How resolved Evidence
Token budget management across 193 KB files P0–P6 priority-based budget assembly with per-tier character caps SystemPromptBuilder.swift
Platform parity (PHP ↔ Swift) for 17 agents Shared agent IDs and knowledgebase; parallel implementations with same prompt structure CLAUDE.md agent inventory, requirements/feature-parity-requirements.md
n8n-first with graceful fallback ITI_Workflow_Adapter (PHP) and N8nWorkflowClient (Swift) try webhook first, fall back to direct Claude N8N-DIFY-GUIDE.md
Duplicate mtpConfig JS variable Identified as defect D001 in smoke testing tests/smoke-test-plan.md
REST validation errors Defect D002 — endpoint parameter validation fixes needed tests/smoke-test-plan.md

Section 7 — AI Tools and Techniques

7.1 AI models and APIs used

Model / API Provider Role in product Integration method
Claude Sonnet 4 (claude-sonnet-4-6) Anthropic Primary advisor model for all 17 agents (WordPress) ITI Shared Library ITI_Claude_API
Claude Haiku 4.5 (claude-haiku-4-5-20251001) Anthropic Lightweight tasks (packing list, title generation) Direct REST in class-mtp-rest.php
Claude (via n8n) Anthropic Orchestrated agent workflows n8n webhook POST /webhook/travelplanner
Tavily Search API Tavily Real-time web search for travel information ITI Shared Library ITI_Tavily_API
Pinecone Pinecone Vector/semantic search over knowledgebase ITI Shared Library ITI_Pinecone_API
text-embedding-3-small OpenAI Embedding model for Pinecone + Dify indexing Dify High Quality indexing mode
Google Places API Google Place search, details, autocomplete iOS GooglePlacesService
Google Earth Engine Google Scenery scoring iOS GEEService
Mapbox Maps (11.20.0) Mapbox Map display and routing iOS/macOS via project.yml

7.2 AI orchestration and tooling

Tool Category Purpose
n8n Workflow orchestration Primary dispatch: classifier → router → specialist AI Agent nodes
Dify Knowledge management Vector retrieval over 183+ files via pgvector, semantic search top_k: 3
ITI Workflow Adapter Integration layer PHP/Swift bridge to n8n with automatic fallback to direct Claude
ITI Shared Library Component library Reusable Claude, Tavily, Pinecone clients + agents + chat handlers

7.3 Prompting techniques used

  • ☑ Budget-controlled dynamic prompt assembly (P0–P6 tiers)
  • ☑ Keyword-based knowledgebase retrieval (deterministic gating)
  • ☑ Semantic vector retrieval (Pinecone + Dify pgvector)
  • ☑ Web search augmentation (Tavily)
  • ☑ Per-agent persona prompts with domain-specific instructions
  • ☑ Safety guardrail injection (medical, diving, legal disclaimers)
  • ☑ Structured output syntax ([[place:]], [[itinerary:{}]])
  • ☑ Conversation history management with token budgets
  • ☑ XML-style prompt structure
  • ☑ Multi-agent routing via n8n classifier

7.4 AI development tools used to build this

Tool How used in build
Cursor IDE Primary development environment for PHP, Swift, and documentation
Claude Code CLAUDE.md stewardship, cross-file refactoring
Antigravity Autonomous test execution, browser QA, visual regression testing — used per global CLAUDE.md tool lane
n8n MCP SDK Workflow creation and management
XcodeGen iOS/macOS project generation from declarative YAML
pytest Integration test suite (6 test suites, live API markers)

Section 8 — Version History and Evolution

8.1 Version timeline

Version / Phase Date Summary of changes Significance
Plugin v1.0.0 / Build 23 Mar 2026 Full WordPress plugin + iOS/macOS app with 17 agents Current release
n8n integration Mar 27, 2026 Docker infrastructure for n8n + Dify orchestration Phase 1 infrastructure
RV + UI features Mar 31, 2026 RV lifestyle features, design system, n8n workflow client Feature expansion
Smoke test Apr 1, 2026 Production validation, 5 defects documented (D001–D005) QA milestone

8.2 Notable pivots or scope changes

  • Evolved from single-agent chatbot to 17-specialist multi-agent system
  • Added n8n/Dify orchestration layer (March 2026) on top of direct Claude calls
  • iOS/macOS native app added alongside original WordPress-only concept
  • FIFA 2026 World Cup planning added as dedicated agent + 26-file KB
  • Browser extension added for save-from-anywhere workflow

8.3 What has been cut or deferred

  • Flyover view (Cesium integration) — in progress per CLAUDE.md
  • Enhanced offline knowledgebase — planned
  • Multi-language support — planned
  • Some data source API keys not configured in all environments
  • AI Engine integration forced to null (API incompatibility)

Section 9 — Product Artifacts

9.1 Design and UX artifacts

Artifact Path Type What it shows
UI Design Spec documentation/UI-DESIGN-SPEC.md Markdown Full UI specification
UI Design Requirements documentation/UI-DESIGN-REQUIREMENTS.md Markdown Design requirements
UI Design Build Plan documentation/UI-DESIGN-BUILD-PLAN.md Markdown Implementation plan

9.2 Documentation artifacts

Document Path Type Status
CLAUDE.md (root) CLAUDE.md Project context Active (March 2026)
CLAUDE.md (product) .claude/CLAUDE.md Detailed product context Active (March 24, 2026)
Detailed requirements documentation/travel-planner-requirements-detailed.md Spec (~2700+ lines) Active
N8N-Dify Guide documentation/N8N-DIFY-GUIDE.md Architecture guide (1018 lines) Active (March 30, 2026)
Test plan documentation/test-plan.md QA plan Active
Smoke test results tests/smoke-test-plan.md Test results (206 lines) April 1, 2026
Shortcode reference documentation/shortcode-reference.html HTML reference Active
Privacy policy documentation/privacy-policy.html Legal Active
Feature parity requirements requirements/feature-parity-requirements.md Cross-platform spec Active
Skills and agents spec requirements/skills-and-agents.md Agent architecture Active
Competitor analysis requirements/competitor-analysis.md Market research Active

9.3 Data and output artifacts

Artifact Path Description
Knowledgebase (193 files) knowledgebase/ Full travel knowledge corpus
FIFA 2026 corpus (26 files) knowledgebase/fifa-2026/ World Cup host city guides + tavily domain filter reference
Browser extension browser-extension/ MV3 Chrome extension for saving places
Plugin install packages plugin-installs/ WordPress plugin zip files
iOS/macOS build archives build/ Xcode build outputs

Section 10 — Product Ideation Story

10.1 Origin of the idea

The product evolved from personal research into travel, expatriation, and RV living (documented in the sibling staging workspace at /Users/peterwesterman/Cursor/ITI/products/my-travelplanner/, now mostly empty as content migrated to this workspace). Initial research sessions using Cursor produced decision matrices and curated reference materials that revealed the potential for an AI-powered travel advisor. [CLAUDE NOTE: inferred from staging workspace content]

10.2 How the market was assessed

Competitor analysis exists in requirements/competitor-analysis.md. First-person travel research informed the product’s multi-domain approach. The decision to support multiple travel personas (expat, RV, scuba, culinary, etc.) came from recognizing that no existing tool covered the full breadth of travel planning decisions. [CLAUDE NOTE: partially inferred]

10.3 The core product bet

That a multi-agent AI system with deep domain knowledge (193 curated files) can provide meaningfully better travel advice than generic AI chatbots or traditional travel planning tools, and that this value is worth delivering across web, mobile, and desktop platforms simultaneously.

10.4 How the idea evolved

Personal lifestyle research (Jan 2026) → structured decision frameworks → concept for AI travel advisor → WordPress chatbot plugin → multi-agent architecture with 17 specialists → iOS/macOS native app with SwiftUI → n8n/Dify orchestration layer → browser extension for save-from-anywhere → FIFA 2026 dedicated agent and corpus.


Section 11 — Lessons and Next Steps

11.1 Current state assessment

What works well: Multi-agent system with 17 specialists; 193-file knowledgebase; n8n/Dify orchestration; iOS/macOS + WordPress parity; budget-controlled prompt assembly Current limitations: 5 defects from smoke testing (D001–D005); Cesium flyover not complete; some API keys not configured; AI Engine integration disabled Estimated completeness: 70% [CLAUDE NOTE: inferred from active tasks list and defect count]

11.2 Visible next steps

  1. Fix defects D001–D005 from smoke testing
  2. Complete Cesium flyover view integration
  3. Configure remaining data source API keys
  4. Implement enhanced offline knowledgebase
  5. Add multi-language support
  6. Complete WordPress ↔ iOS feature parity
  7. Expand knowledgebase coverage

11.3 Lessons learned

_Manual input required — this section cannot be populated automatically._


Section 12 — Claude Code Validation Checklist

  • ☑ Every placeholder has been replaced or marked NOT FOUND
  • ☑ All externally-sourced competitive data is marked with ⚡
  • ☑ All inferences are marked with [CLAUDE NOTE]
  • ☑ Version history is derived from actual git log where available
  • ☑ Knowledge system paths reflect real directory structure
  • ☑ AI tools are confirmed from code/config, not guessed
  • ☑ Section 11.3 is left blank for manual input
  • ☑ Document header shows today’s date and files examined

Sources Examined

File / Path What it contributed
CLAUDE.md (root) Product overview, agent inventory, KB structure, development status, n8n/Dify integration ref
.claude/CLAUDE.md Detailed platform specs, features, architecture, file locations, build commands
my-travelplanner/my-travelplanner.php Plugin header (v1.0.0), bootstrap logic, 17 agent class loads, DB tables
my-travelplanner/composer.json Package metadata, dev dependency (wordpress-stubs 6.7)
MyTravelPlanner/project.yml iOS/macOS build config: Swift 5.9, Mapbox 11.20.0, iOS 17.0, macOS 14.0, Build 23
documentation/N8N-DIFY-GUIDE.md n8n/Dify architecture (1018 lines), webhook config, Dify retrieval params
requirements/skills-and-agents.md Model preferences (Opus 4.6 vs Sonnet 4.6), orchestration expectations, bug/feature backlog
MyTravelPlanner/MyTravelPlanner/Services/SystemPromptBuilder.swift P0–P6 budget assembly (548 lines), place hotlink syntax, agent personas
.claude/settings.json Permission baselines (bash, web fetch domains)
tests/pytest.ini Test suite structure (6 suites), markers
tests/requirements.txt Python test dependencies (pytest, httpx, requests)
browser-extension/manifest.json “Save Places” extension v1.0.0, content scripts for travel sites
tests/smoke-test-plan.md Smoke test results (Apr 1, 2026), defects D001–D005
Git log (13 commits) Full commit history with dates and scopes

Addendum — April 2026 Competitive Landscape and Roadmap Update

1. Industry Context

The AI-powered travel planning space has undergone a structural shift since My TravelPlanner’s initial build in March 2026. The vibe coding ecosystem — now a $9.4 billion funded category with 41% of global code AI-generated — has lowered barriers to entry for travel app competitors. Tools like Bolt.new and Lovable mean a motivated non-developer can stand up a basic AI trip planner in days. This is the environment My TravelPlanner now competes in: not a scarcity of AI travel tools, but a surplus.

More consequentially, Model Context Protocol (MCP) has emerged as a new distribution channel for travel services. Expedia, Booking.com, and airlines have published MCP servers. Startups like aitrips.io and Lambus now let travelers plan trips inside Claude, ChatGPT, or Gemini — never visiting a dedicated travel app at all. This is a fundamental channel shift: the “app as destination” model is being supplemented (and in some cases replaced) by “agent as storefront.” MTP must decide whether to fight this pattern or adopt it.

Meanwhile, frontier LLM convergence means “AI-powered” is no longer a differentiator. Claude 4, GPT-5, and Gemini 2.5 score within 10% of each other on reasoning benchmarks. Every travel app can integrate the same models at similar cost. The differentiator is no longer the model — it’s the domain expertise, orchestration architecture, and curated knowledge embedded in the product.

2. Competitive Landscape Changes

New Entrants (Since March 2026)

Competitor Type Key Threat
Travo AI-first mobile trip planner Free AI itinerary generation in seconds; native iOS/Android with free offline access
Layla.ai AI trip planner (web) Complete itineraries with flights, hotels, activities, dining; $49.99/yr
aitrips.io MCP-native trip persistence Plans trips inside any AI assistant with persistent storage
Lambus AI Connector multi-LLM planner Works with ChatGPT, Claude, Gemini via MCP
Tripomatic 26 Full-featured planner refresh AI assistant, 50M places, offline maps, multi-modal routing
Expedia MCP Server Infrastructure/distribution Open-source MCP server for hotels, flights, activities, car rentals

Features Competitors Have Added

Feature Who Added It Impact on MTP
MCP server integration aitrips.io, Lambus, Expedia New distribution channel MTP doesn’t participate in
Free AI itinerary generation Travo (free), Layla ($49/yr) Raises baseline expectation for AI trip planning
Calendar event import Tripsy 3.8.0 Table-stakes feature MTP lacks
Multi-modal routing (bike, taxi, scooter, transit) Tripomatic 26 Extends beyond driving/walking that MTP supports
Offline maps with auto-expiry Tripomatic 26 Smart offline approach MTP can learn from

Eroded vs. Sustained Differentiators

Differentiator Status
“AI-powered conversational trip planning” Eroded — Travo, Layla, Wanderlog, Tripomatic, and raw LLMs all offer this
“Cross-platform delivery” Eroded — Tripomatic 26 now has iOS + Android + web with sync
17 specialist AI agents with domain routing Still unique — no competitor offers scuba, expat, RV, culinary specialist routing
193-file curated RAG knowledgebase Still unique — competitors use generic LLM knowledge or partner APIs
Multiple itineraries per trip Still unique
Kanban board for trip planning Still unique
Google Earth Engine scenery scoring Still unique
FIFA 2026 dedicated advisor + 26-file corpus Time-limited — value peaks June-July 2026

3. Our Competitive Response: Product Roadmap

The roadmap is organized into four tiers reflecting urgency and strategic value.

Tier 1 — Critical (4-6 weeks): Close quality gaps and establish a market-ready baseline. Fix the 5 smoke-test defects (D001-D005). Connect the disconnected TodayItineraryWidget. Ship table-stakes features users now expect: public shareable trip links, PDF export, iCal export, and offline data access. Most importantly, build full AI itinerary generation from a natural language prompt — orchestrating across the 17 specialist agents to produce itineraries richer than any single-model competitor can generate.

Tier 2 — High Value (6-8 weeks): Capture the MCP distribution opportunity by publishing MTP’s 17 specialist agents as an MCP server — enabling Claude, ChatGPT, and Gemini users to access scuba, RV, expat, and culinary travel advisors from inside their preferred AI tool. No travel app currently exposes specialist AI agents via MCP. Also: route optimization (TSP solver), collaborative editing, weather in itinerary, and calendar event import.

Tier 3 — Strategic (8-12 weeks): Deepen the moat with email booking auto-parse, expat scouting trip dashboard, hotel price comparison, flight search integration, community destination guides, and major knowledgebase expansion (Asia, Middle East, Oceania — ~40-50 new files).

Tier 4 — Exploratory: Apple Watch companion, full web app, Android app, adaptive real-time itinerary adjustments, and multi-language support. Deferred until core experience is polished.

Deprioritized: Cesium flyover view (low user-facing value vs. effort), AI Engine integration (unresolved API incompatibility — direct Claude API + n8n is the strategic path), and full WordPress-iOS feature parity (iOS/macOS is the primary platform).

4. New Capabilities Added Since Last Build

The following Skills have been added to the ITI skill library since the showcase was written, directly supporting roadmap execution:

Skill Relevance to My TravelPlanner
mcp-server-development Core infrastructure for T2-01: publishing 17 specialist agents as an MCP server
mapbox-offline-packs Offline tile download management for T1-07 offline data access
icalendar-ics-generation Standards-compliant .ics feed generation for T1-06 calendar export
tsp-route-optimization TSP solver and Mapbox Optimization API integration for T2-02 route optimization
email-parsing-travel-bookings Parsing airline/hotel/restaurant confirmations for T3-01 email auto-parse
scouting-trip-planning Structured evaluation trip methodology for T3-02 expat scouting dashboard
fifa-2026-travel FIFA 2026 World Cup travel advisory for the dedicated FIFA agent’s time-limited window

5. Honest Assessment

Strengths: The 17-specialist multi-agent architecture with 193-file curated RAG remains genuinely unique in the travel planning market. No competitor orchestrates across domain-specific advisors (scuba, expat, RV, culinary) — they all use single generic LLMs. The n8n/Dify orchestration layer provides a flexible integration backbone. The FIFA 2026 corpus is timely and irreplaceable for the tournament window.

Gaps: Five smoke-test defects remain open. Table-stakes features (shareable links, offline access, PDF export, calendar sync) that every competitor offers are missing. The TodayItineraryWidget is disconnected. Several data source adapters remain stubs. Estimated completeness is 65-70%.

What we’re watching: The MCP distribution shift is the most consequential change. If travelers routinely plan trips inside Claude or ChatGPT using MCP-connected travel services, apps that don’t participate in this channel risk irrelevance. The T2-01 MCP server is the single most strategic item on the roadmap. We’re also watching Apple’s Siri overhaul (iOS 27, WWDC June 2026) — if Siri gains meaningful travel planning capability via third-party AI extensions, the iOS app’s competitive position changes.

Portfolio context: My TravelPlanner demonstrates ITI’s ability to build and orchestrate complex multi-agent AI systems across platforms (WordPress, iOS, macOS). The multi-agent routing, budget-controlled prompt assembly, and n8n workflow patterns are directly reusable in consulting engagements. The product is portfolio evidence of AI systems architecture — not a consumer play competing on marketing spend.