PathHub AI Engine

Our planning engine: several steps instead of one prompt

PathHub AI is no generic chatbot wrapper. Our model-agnostic engine picks the best available model for every task. The intelligence lives in the architecture around it: RAG-based project context, specialist routing, persistent memory and Agentic AI that actually gets work done.

Model-Agnostic Engine RAG Context Specialist Routing Learns over time GDPR · EU servers
Start free How it works
chunks_matched: 7
router → planner
memory.recall()
PathHub AI Engine · live
v4.7.2
1
Input & Intent
Parse user query · classify intent
parse
2
Context Layer (RAG)
Project data + documents + memory
retrieve
3
Specialist Routing
Pick best engine + prompt architecture
route
4
Verification
Output checks · consistency · accuracy
verify
The Mechanics

How our AI actually works

Four layers tailoring every response to your project, from semantic search of your data to verified output.

Learning AI

Your AI gets sharper with every project

PathHub AI learns from your projects, without training your data into a model. While you work, PathHub records how things actually go: which phases take longer than planned, where budgets deviate, what gets delayed. When a project is completed, this turns into lessons. Together with your company profile and your documents (RAG), these experience values feed every new plan, every recommendation and every answer of the assistant. The more projects you complete, the more precisely PathHub plans for your company.

Plan/Actual
measured for every phase, every budget and every delay
0
model training on your data · servers in Germany
Move the timeline
Week 1 Week 6 Week 12+
Agentic AI

Your AI colleagues that actually do the work

No more "ask the chatbot". Our Agentic AI runs complex tasks fully autonomously: analyzes, prioritizes, writes, applies recommendations, and delivers finished results. One click, one result, zero prompt engineering.

Agents · active
6 tasks · live
Agent spotlight
Weekly Status Report Agent
Analyzes your full project state and produces a structured weekly status: progress, budget, risks and next steps.
    Common tasks
    All AI Features

    AI woven into every part of the app

    PathHub AI is not an isolated chat box. The AI is built into every workflow, from the first project idea to the final report.

    AI Project Planning
    Describe your idea, get a complete plan with phases, tasks, budget, risks & stakeholders.
    Custom prompts & PM style configurable
    Automatic compliance check (104+ rules)
    All plans
    AI Recommendations
    Prioritized, actionable recommendations with impact rating, applied to your plan with one click.
    High · Medium · Low priority
    Agentic Apply (1-click execution)
    Business plan
    AI Assistant
    Your project advisor: knows every phase, every task, every budget line. Answers based on your real project context.
    Quick shortcuts right inside the project
    References your workspace documents
    Free · Pro · Max
    Agentic AI
    Autonomous agents for status reports, weekly plans, stakeholder updates, risk scans: finished results, one click.
    6 built-in agents + free-text prompts
    Documents persist in your workspace
    Business plan
    AI Toolbox · 11 Tools
    RACI, Eisenhower, Risk Matrix, Stakeholder Map, SWOT, WBS, Gantt, Cost Compare & more: all interactive, all from your data.
    Inline edit, drag-drop, auto-save
    Export as PNG from any tool
    Business plan
    Compliance Library
    95+ regulatory rules (GDPR, EAA, ISO, IT-SG, …), auto-detected in your project context and prioritized.
    Criteria · actions · penalties · sources
    Filterable by industry
    All plans
    At the core
    We don't rely on one model. We pick the right specialist for each task.
    G
    GPT-6 Sol
    Structured outputs: text polishing, subtasks, matrix & resource tools, format checks on plans.
    L
    GPT-6 Luna
    The reasoning model for the demanding work: project planning, recommendations, risk analysis, project check, agents & documents.
    C
    Claude Haiku 5.5
    AI assistant in chat: answers questions with reference to your plans, documents and workspace context.
    Models are interchangeable Our engine is model-agnostic. The magic lives in the architecture around them: RAG pipeline, prompt specialization, memory layer and verification. When a model gets better, PathHub gets better automatically: no migration, no effort on your side.
    In the app

    How the AI works inside your Path

    Not a chat island next to the project, but insights, answers and tools right in the plan.

    AI insights in the Path

    Recommendations with reasons that you apply or look at in detail first.

    AI insights in the Path

    AI chat with project context

    The assistant knows the Path's phases, tasks, budget and risks and answers specifically.

    AI chat with project context

    AI toolbox

    Interactive tools such as a RACI matrix, risk matrix or stakeholder map, filled from the project.

    AI toolbox

    Real views from a sample Path (launching a SaaS product module), not mock-ups.

    Continuous development

    The AI engine keeps getting better

    New capabilities, deeper integrations, smarter learning: a glimpse of what's on our roadmap.

    Shipped March 2026
    11 interactive AI tools
    All PM tools (RACI, Eisenhower, Risk, Stakeholder, WBS …) now with inline edit, drag-drop and auto-save.
    Shipped March 2026
    Compliance library expanded to 104+ rules
    Every rule enriched with criteria, actions, penalties and source links: automatically applied in your project context.
    Shipped April 2026
    Custom prompts & PM style
    New fields in the project setup: choose your PM style (Scrum, Kanban, Waterfall …) and add a free-text custom prompt the AI follows precisely.
    Shipped April 2026
    Agentic AI v2 with persistent memory
    All agent outputs are stored, free-text prompts for custom document types, expanded context usage.
    Shipped June 2026
    Documents as AI context
    Uploaded documents are chunked, embedded and retrieved per request to match the project at hand: the foundation of the RAG pipeline.
    Shipped July 2026
    Chat answers word by word
    The AI assistant streams its response instead of delivering it in one block. The status phases shown map to the real processing steps.
    Shipped July 2026
    Compliance rules with criteria and actions
    Every rule gained checkable criteria, concrete steps and the consequences of non-compliance, instead of just naming the statute.
    Shipped September 2026
    New reasoning model in production
    From September, plan generation ran on GPT-5.6 Luna: more depth on nested dependencies and more consistent plans across all sections.
    Live now September 2026
    Deeper workspace RAG integration
    Documents from Google Drive and SharePoint are synced, chunked and embedded automatically: the AI assistant pulls them in as context on every answer, with nothing to upload manually.
    Live now September 2026
    Parallel phases and tasks
    Plan generation now schedules in two dedicated steps: a scheduler decides, with a content-based justification, which phases may overlap (at most half, never before their predecessor) before budget and risks build on that real timeline. Within each phase, tasks start in parallel and get predecessors instead of a chain of equal blocks. Milestones sit on the actual phase ends.
    Live now September 2026
    Multi-agent orchestration: the project check
    Three specialised agents work as a chain: the risk scanner delivers evidence-backed findings, the measure planner derives concrete changes from them, the budget checker weighs them against the budget and holds a veto. A coordinator validates every change against the real project before you apply it with one click (or automatically, with undo).
    Live now October 2026
    New model generation: GPT-6 and Claude Haiku 5.5
    Plans, agents and the project check run on GPT-6 Luna, scheduling and helper tasks on GPT-6 Sol, the AI assistant in chat on Claude Haiku 5.5. Tested on real projects first: plans keep deadlines and task rules more precisely, tasks reliably get dependencies, and the chat answers faster and more precisely.
    Coming Q4 2026
    Proactive AI: suggestions without prompting
    The AI watches your project and suggests adjustments proactively. Before issues arise.
    Privacy & trust

    Data protection under GDPR

    Carefully chosen providers, minimal data sharing, full transparency.

    Servers in Germany

    All project data is stored on a Hetzner server in Falkenstein, Germany. All subprocessors are listed in the Trust Center.

    No training on your data

    Neither OpenAI nor Anthropic use API requests for training. What you send stays yours.

    Minimal data sharing

    Only project data. Never: email, name, password, payment info or IP address.

    You always decide

    Every AI output is a suggestion. Accept, edit, discard: full control stays with you.

    Data flow at a glance

    What goes where, and how long it is stored.

    Data Where it goes Retention
    Project data, documents & workspace context Hetzner servers, Falkenstein (EU) Until you delete it
    Per AI request: relevant project context (RAG) Anthropic / OpenAI API: under a DPA, no training Provider-side max. 30 days (abuse monitoring), then deleted
    Name, email, password, payment data, IP Never sent to AI providers —
    Full privacy policy, section 5 →

    See for yourself: in 2 minutes.

    Create your first Path for free. No credit card, no hidden costs: instant AI-generated project plans.

    Start free