🔒 Enterprise Feature

Set a goal. Watch a simulated user try to achieve it. Learn where they give up.

cognitive_journey_autonomous runs a complete user journey with AI making all navigation decisions. No orchestration needed — just specify a persona, goal, and starting URL. The AI experiences your site the way a real user would, complete with frustration, confusion, and the moment they decide to leave.

Quick Start

{
  "persona": "first-timer",
  "startUrl": "https://example.com",
  "goal": "Sign up for a free trial",
  "maxSteps": 30,
  "vision": true
}

What happens:

AI loads the persona's cognitive profile (26 traits)
AI perceives the page as that persona would
AI decides what action to take based on persona's characteristics
After each action, cognitive state updates (patience, confusion, frustration)
If any metric crosses abandonment threshold, journey ends
Final report shows exactly where and why they gave up

Parameters

Parameter	Type	Required	Default	Description
`persona`	string	Yes	—	Persona ID (e.g., `first-timer`, `elderly-user`, `cognitive-adhd`)
`startUrl`	string	Yes	—	URL where the journey begins
`goal`	string	Yes	—	What the persona is trying to accomplish
`maxSteps`	number	No	50	Maximum actions before stopping
`maxTime`	number	No	300	Maximum time in seconds
`vision`	boolean	No	false	Send screenshots to AI for visual analysis
`headless`	boolean	No	false	Run without visible browser window
`location`	object	No	—	Override persona's timezone, locale, geolocation

Location Parameter

Override the persona's default location to simulate users from specific regions:

{
  "persona": "elderly-user",
  "startUrl": "https://bank.example.com",
  "goal": "Transfer money",
  "location": {
    "timezone": "America/New_York",
    "locale": "en-US",
    "geolocation": { "latitude": 25.7617, "longitude": -80.1918 }
  }
}

Property	Description	Example
`timezone`	IANA timezone ID	`America/Los_Angeles`, `Europe/London`
`locale`	BCP 47 locale	`en-US`, `de-DE`, `ja-JP`
`geolocation`	Lat/lng coordinates	`{ "latitude": 40.7128, "longitude": -74.0060 }`

Priority: Explicit location > Persona default > CBrowser default (America/Denver)

Vision Mode

When vision: true, the AI receives screenshots at each step, enabling:

Understanding visual layouts that aren't captured in DOM text
Navigating dropdown menus that require hover states
Reading non-semantic content (images with text, icons without labels)
Perceiving visual hierarchy the way humans do

Vision mode consumes more API credits but dramatically improves accuracy on complex sites.

Use Cases

1. Onboarding Flow Validation

Goal: Verify that new users can complete signup without confusion.

{
  "persona": "first-timer",
  "startUrl": "https://app.example.com",
  "goal": "Create an account and reach the dashboard",
  "vision": true
}

What you learn: Where first-timers get lost, which steps deplete patience, what causes confusion.

2. Accessibility Impact Testing

Goal: Understand how accessibility issues affect real task completion.

{
  "persona": "motor-tremor",
  "startUrl": "https://example.com/checkout",
  "goal": "Complete a purchase",
  "vision": true,
  "maxSteps": 40
}

What you learn: Which interactions are impossible with motor impairment, where the user gives up.

3. Mobile UX Validation

Goal: Test if the mobile experience works for impatient users.

{
  "persona": "impatient-user",
  "startUrl": "https://m.example.com",
  "goal": "Find pricing information"
}

What you learn: Whether navigation is fast enough for users with low patience tolerance.

4. Competitive Research

Goal: See how different user types experience a competitor's site.

{
  "persona": "enterprise-buyer",
  "startUrl": "https://competitor.com",
  "goal": "Request a demo",
  "vision": true
}

What you learn: Friction points in their conversion funnel that you can avoid.

How It Works

Decision-Making Process

At each step, the AI:

Perceives the page — What elements are visible? What can be interacted with?
Applies persona traits — A curious user explores sidebars. An impatient user seeks the fastest path.
Evaluates options — Rank possible actions by likelihood of achieving goal
Makes a decision — Choose action based on persona's risk tolerance, comprehension, etc.
Updates state — Adjust patience, confusion, frustration based on what happened

Abandonment Logic

The persona abandons when:

Trigger	Threshold	Inner Monologue
Patience depleted	< 0.1	"This is taking forever, I'm done."
Too confused	> 0.8 for 30s	"I have no idea what to do here."
Too frustrated	> 0.85	"This is ridiculous."
No progress	10+ steps, < 0.1 progress	"I'm not getting anywhere."
Stuck in loop	Same page 3+ times	"I keep ending up back here."

Output

{
  "journeyId": "journey_abc123",
  "persona": "first-timer",
  "goal": "Sign up for a free trial",
  "outcome": "abandoned",
  "goalAchieved": false,
  "totalSteps": 18,
  "totalTime": 142,
  "abandonmentPoint": {
    "url": "https://example.com/signup/step-2",
    "action": "Attempted to fill 'Company Size' dropdown",
    "reason": "confusion_threshold_exceeded",
    "innerMonologue": "I don't know what to pick here. The options don't match my situation."
  },
  "finalState": {
    "patience": 0.23,
    "confusion": 0.82,
    "frustration": 0.67
  },
  "frictionPoints": [
    {
      "step": 5,
      "url": "https://example.com/signup",
      "issue": "Password requirements unclear until submission",
      "patienceImpact": -0.12
    },
    {
      "step": 12,
      "url": "https://example.com/signup/step-2",
      "issue": "Dropdown options don't include 'Individual/Freelancer'",
      "confusionImpact": 0.25
    }
  ],
  "decisionTrace": [
    {
      "step": 1,
      "thought": "I need to find where to sign up",
      "action": "click",
      "target": "Get Started button",
      "confidence": 0.9
    }
    // ... full trace of every decision
  ],
  "screenshots": [
    {
      "step": 12,
      "label": "abandonment_point",
      "base64": "..."
    }
  ]
}

Tips & Best Practices

Choose the Right Persona

If you want to test...	Use this persona
General usability	`first-timer`
Speed and efficiency	`power-user` or `impatient-user`
Mobile experience	`mobile-user`
Accessibility	`motor-tremor`, `low-vision`, `cognitive-adhd`
Enterprise sales funnel	`enterprise-buyer`
Consumer checkout	`impulse-shopper` or `price-researcher`

Set Appropriate Limits

Simple flows: maxSteps: 20, maxTime: 120
Complex flows: maxSteps: 50, maxTime: 300
Exploratory: maxSteps: 100, maxTime: 600

Interpret Results

Goal achieved quickly → Good UX for this persona
Goal achieved with high frustration → Success but poor experience
Abandoned early → Critical friction point
Abandoned late → Death by a thousand cuts

Why This Requires Enterprise

cognitive_journey_autonomous runs AI decision-making at every step, which:

Consumes Anthropic API credits — Each step requires AI inference
Requires secure API key storage — Keys must be managed properly
Uses significant compute — Vision mode processes screenshots

Demo users can still run cognitive journeys using the bridge workflow (cognitive_journey_init + cognitive_journey_update_state), where Claude orchestrates each step manually.

Contact for Enterprise →

Related Tools

cognitive_journey_init — Start orchestrated journeys
compare_personas — Compare multiple personas
empathy_audit — Accessibility-focused simulation

Last updated: v18.9.0 — Added location parameters (timezone, locale, geolocation)

Tool Cognitive Journey Autonomous