Introduction

Overview

Praxiom AI is the copilot for product managers. It transforms raw user research — interviews, support tickets, feedback surveys — into structured insights, prioritized recommendations, and polished documents. Beyond the core analysis pipeline, Praxiom provides a full PM IDE with structured document blocks, execution ticket generation, conversation intelligence, and integrations with GitHub, Linear, Jira, and Google Drive.

The Product Loop

Praxiom follows a six-stage Product Loop:

1. Upload — Bring in your raw research (PDFs, audio, CSVs, URLs, images)
2. Synthesize — AI extracts patterns, themes, and evidence-backed insights
3. Recommend — Get prioritized, evidence-backed feature recommendations scored by impact and effort
4. Draft — Generate PRDs, specs, memos, and more grounded in your research
5. Structure — Break documents into typed blocks (problems, personas, user stories, metrics) and refine each with AI-powered expand, challenge, and simulate actions
6. Execute — Generate scoped tickets from document blocks and push them to GitHub, Linear, or Jira

Navigating the Interface

The sidebar follows a pipeline-centric layout with a vertical connector line through the four core stages:

1. Research — Upload and manage research sources
2. Insights — View synthesized findings and patterns
3. Recommendations — Browse prioritized feature recommendations
4. Documents — Access drafted PRDs, specs, and memos

Below the pipeline, a collapsible Tools section provides access to:

• Knowledge Graph — Visual entity relationship map
• PM IDE — Structured document editor with typed blocks
• Missions — Multi-agent command center

The sidebar collapses to icon-only mode for more screen space. Your workspace switcher, usage badge, and account settings live at the bottom.

Core Capabilities

AI Agent

A conversational copilot powered by Claude (Haiku, Sonnet, or Opus) with configurable effort levels. The agent adapts its persona based on your task — research analyst, strategy advisor, writer, PM agent, or explorer. Conversations maintain intelligence across sessions with purpose tracking, working state, and continuity chains. Attach files directly in chat with vision support for images and documents.

Deep Research

Use the /research command to trigger deep research with a depth selector (quick, standard, deep) and upfront cost estimate before execution.

Document Blocks & PM IDE

Documents are composed of typed blocks — problem, persona, solution, metrics, user_story, edge_case, risk, timeline, and free_text. Each block can be expanded, challenged, or simulated by the AI. The PM IDE provides a structured editing environment for block-based documents.

Execution Bridge

Turn document blocks into actionable engineering tickets. Generate tickets with titles, descriptions, acceptance criteria, effort estimates, and priority levels, then push them to GitHub, Linear, or Jira in bulk.

Knowledge Graph

Visualize the relationships between research sources, insights, recommendations, and documents. LLM-powered entity extraction identifies people, pain points, features, workflows, themes, decisions, and competitors from your research, creating a rich semantic network.

Dashboard Quality

Monitor workspace health with the Quality Overview — contract pass rates, complexity distribution, RQS trends, and entropy scanning. Quality badges (RQS, RecQS, DocQS) appear throughout the UI.

Overnight Research Cycles

Power plan users can enable autonomous nightly experiments that explore different research angles, score results by quality, and deliver a morning digest of new findings.

Skills

Extend the agent with installable skills from the official library or create custom skills with your own instructions and trigger keywords.

Integrations

Connect GitHub, Linear, Jira, and Google Drive to close the loop between product decisions and engineering execution. Bidirectional sync keeps your workspace aligned with external tools. Beyond OAuth-based push, the agent gets direct read tools on every integration — 7 GitHub tools for grounding in code, 9 Linear tools for querying issues and cycles, 8 Google Drive tools for searching and importing files. See GitHub, Linear, Google Drive, and Jira for the full tool catalogs.

Human-in-the-Loop (Pending Actions)

High-stakes writes — saving recommendations, creating Linear issues, updating Drive files — are queued for human approval before they land. Every @gated tool produces a pending action that you can approve, edit, or reject. Batch approval groups actions from a single mission so you can ship an entire plan in one click. See Pending Actions and the API reference.

Agent Harness

Every agent run is wrapped by a post-flight quality system: an algorithmic contract check, an independent Haiku-based verifier, mid-stream alerts at character checkpoints, retry loops with healing, and workspace-profile-aware plan selection. The harness emits 17+ SSE events so frontends can surface quality signal in real time. See Agent Harness.

Validation Kit

For idea-stage product work, the Validation Kit generates a linked triad of documents — Assumption Map (what you're betting on), Research Plan (how to test it), and Interview Guide (the questions) — as a single background task with live progress over SSE. See Validation Kit.

Verification & Trust

Every AI response can be verified with citation checks, cross-source validation, severity analysis, and density scoring. Trust badges provide at-a-glance confidence signals.

Credit-Based Billing

AI operations are billed using a cost-proportional credit system (1 credit = $0.08 USD). Three paid plans (Pro, Growth, Power) offer increasing credits, feature gates, and capabilities. Subscriptions are user-scoped — one plan covers all your workspaces.

What's Next

Get up and running in 5 minutes with the Quickstart guide.