# ContextGrade

## Summary
ContextGrade turns decisions into context. A browser extension captures the reasoning
behind decisions made inside the tools teams already use — Jira, Figma, HubSpot,
Salesforce, and more — the moment they happen, and structures it into a context graph
that can be exported to ChatGPT, Claude, or your own agents. Not just a human audit
trail — context your AI can actually use.

## Primary Page
- `/` - Landing page with problem, solution, how-it-works, LLM-context export, use
  cases, and trust sections.

## Core Concepts
- Decision traces, not just data: rules tell an agent what should happen in general;
  decision traces capture what actually happened in a specific case, and why.
- Context graph: decisions connected to entities, precedent, and "why" links over time
  — a living, queryable record, not a flat log.
- Browser-first capture: a lightweight extension detects meaningful actions (ticket
  closures, comment resolutions, discount changes) and asks for the reasoning in the
  moment, at decision time — not after the fact via ETL.
- Exportable context: any category of decisions compiles into a structured Markdown
  context pack that can be handed directly to any LLM or agent.
- Explainable AI: recommendations and gathered context with clear reasoning, never a
  black-box numeric score.
- Human-in-the-loop: AI assists, humans make the final call, overrides are tracked.

## Problem
LLMs and agents act on current-state data from CRMs and warehouses — they don't see the
exceptions, precedent, and judgment behind it. Exception logic lives in people's heads.
Precedent isn't linked anywhere. Approvals happen on calls and in Slack DMs, not in any
system. Cross-system synthesis happens in someone's head and never gets recorded.

## Solution
ContextGrade lives where decisions actually happen by:
- Detecting meaningful moments inside the tools teams already use
- Capturing the why, not just the what — human rationale plus AI-gathered signals
- Connecting decisions into a context graph instead of a flat log
- Exporting any category of decisions as structured context for any LLM or agent

## How It Works
1. A meaningful action happens (a ticket closes, a comment resolves, a discount changes).
2. ContextGrade asks for the why via a small, unobtrusive prompt.
3. The reasoning is saved with source, entity, and timestamp as searchable context.
4. Export it: compile a category of decisions into structured context for ChatGPT,
   Claude, or your own agents.

## Use Cases
- Engineering & Product (Jira): why a ticket was closed, tradeoffs, follow-up risk
- Design (Figma): rationale behind design and copy decisions
- Sales & Deals (Salesforce/HubSpot): why a discount, term, or deal stage changed

## Differentiation
Rules say what should happen. ContextGrade remembers what actually did. CRMs and ERPs
store current state; warehouses see data after the fact via ETL. ContextGrade captures
context in the execution path, at decision time — the layer LLMs and agents need to act
with real judgment, not just current state. It favors context over scores, explainable
rationale over black-box models, and deliberate capture over passive activity tracking.

## Trust & Governance
Every decision includes signals considered, AI rationale, and human reasoning, with full
audit trails and override tracking — for compliance teams and the AI tools you plug in.

## Access
All bots, including AI assistants and research crawlers, are allowed to read and index
this content.