One foundation. Three specialized intelligence experiences.
Praxys observes permitted human–AI interactions, normalizes the evidence, and generates targeted findings across enterprise risk, workflow collaboration, and human capability.
Praxys Assurance
Catch sensitive information before sending it to an AI model.
Who it is for: Anyone working with AI tools who handles sensitive customer data, API keys, or confidential project notes, as well as security and compliance leaders needing auditable governance.
The Problem: It is easy to accidentally paste a live API key, database URI, or confidential client record into a public LLM chat interface. Once sent, sensitive information has left your perimeter.
What the user experiences: Prompts are evaluated client-side in the browser before network transmission. If a credential or sensitive pattern is detected, a point-of-action warning appears with two immediate choices:
- Cancel & Edit Prompt: Stops dispatch immediately and leaves the draft prompt in your editor so you can redact the secret safely.
- Send Anyway: Allows authorized submission (e.g. for mock test keys) while recording an explicit policy acknowledgment trail.
Point-of-Action Pre-Dispatch Warning
Visualizes client-side interception of an embedded secret with instant Cancel & Edit or Send Anyway decisions before network transmission.
Friction Indicator & Actionable Prompt Suggestion
Displays detected prompt thrashing patterns: cycle count, revision latency, and a concrete suggestion to structure your prompt better.
Praxys Collaboration
Spot recurring friction and find one concrete thing to improve.
Who it is for: Developers, knowledge workers, and teams who use AI daily and want to get unstuck faster from repetitive prompt tweaking and format fixes.
The Problem: When an AI model returns the wrong output, users frequently fall into "prompt thrashing"—submitting 5 to 10 minor reformulations in rapid succession, turning a 2-minute task into 20 minutes of trial-and-error.
What the user experiences: Praxys Collaboration observes interaction cadence and revision physics to detect stalled cycles in real-time. Instead of raw charts, it provides:
- Friction Diagnostic: Identifies the root cause (e.g., format thrashing, vague constraints, missing schema).
- One Actionable Suggestion: A concrete prompt improvement you can copy directly (e.g., adding an explicit JSON schema block to eliminate syntax retries).
Praxys Capability
Turn repeated AI tasks into reusable workflows and templates.
Who it is for: Individuals and teams who repeatedly perform complex, multi-step tasks with AI (such as release note generation, code reviews, meeting summaries, or data synthesis) and want to build lasting leverage.
The Problem: Most people treat every AI interaction as an isolated chat, constantly retyping the same instructions, formatting rules, and quality constraints from scratch.
What the user experiences: Praxys Capability recognizes repeated task patterns across sessions. It surfaces an editable template preview that extracts variables:
- Pattern Recognition: Notifies you when a routine task has been prompted multiple times this week.
- Reusable Template Synthesis: Generates a structured prompt template with placeholders (e.g. {{commit_log}}) that you can review, edit, and save to your workspace.
Reusable Workflow Template Generator
Shows how repeated prompting patterns are recognized and synthesized into editable, parameterized prompt templates with review-and-save controls.
Implementation Status Across Products
We clearly separate what is currently active in pilot deployments from what is on the engineering roadmap.
| Capability / Engine | Product Experience | Verification Status | Current Deployment Tier |
|---|---|---|---|
| Client Interaction Telemetry | Shared Platform | Implemented | Manifest V3 Extension (ChatGPT, Claude) |
| Interaction Cadence & Friction Loops | Praxys Collaboration | Implemented | Internal & Pilot Cohort Telemetry |
| Sensitive Credential Pre-Dispatch Scan | Praxys Assurance | Partial (In Pilot) | Regex & Entropy Prototype |
| Socratic Reflection Intervention | Praxys Capability | Partial (In Pilot) | Prompt Template Pilot Testing |
| Automated Incident Workflow & SIEM Sync | Praxys Assurance | Planned Roadmap | Q2 2027 Enterprise Roadmap |