Version: v2.1

📚 Parse Developer Hub

Security architecture, reasoning engine documentation, and integration guides.

👋 Welcome to Parse

Parse is a Twin-Engine Growth Platform that connects to your commercial data layer to detect Risks (Revenue Leakage) and Opportunities (Expansion). Unlike traditional BI tools that require SQL and maintenance, Parse uses pre-trained autonomous agents to scan your stack 24/7.

This documentation covers our security architecture, the "Multi-Node LangGraph" reasoning engine, and integration guides.

🚀 1. Getting Started

Quickstart: Connect in 3 Minutes

Parse requires no code to get started. We use standard OAuth flows and API keys to establish read-only connections.

  1. 1

    Create an Account

    Sign up with your corporate email.

  2. 2

    Connect Sources

    Navigate to Settings → Integrations. We recommend connecting at least three distinct layers:

    • • CRM: (e.g., Salesforce/HubSpot)
    • • Financial: (e.g., Stripe/Chargebee)
    • • Operational: (e.g., Jira/GitHub)
  3. 3

    Initial Scan

    The "Standard 100" agents will immediately begin scanning historical data. Initial insights typically appear within 30-90 minutes.

The "Commercial Truth" Model

Data in modern SaaS organizations is siloed:

  • HubSpot knows what was sold.
  • Chargebee knows what is invoiced.
  • The Platform knows what is actually used.

Each system is correct in isolation, but together they tell different stories. Parse connects these layers to create a unified "Commercial Truth," identifying gaps where revenue leaks or churn hides.

🔒 2. Security & Privacy

Philosophy: We are a "Glass Box." You should know exactly how your data is accessed, processed, and stored.

Data Security & Architecture

We have built a security-first architecture designed for enterprise requirements.

Encryption

End-to-end encryption in transit (TLS 1.3) and at rest, protected with AES-256.

Access Control

Role-based access (RBAC), comprehensive audit logs, and read-only data handling by default.

Control

You retain full ownership. You can request full data deletion at any time.

SOC 2 & Compliance Process

While we operate on a security-first architecture, we have not yet prioritized a formal SOC 2 audit (planned for next year). To support your vendor review process immediately, we offer:

  • Security Questionnaires: We are happy to complete your internal security assessments.
  • DPAs: Standard Data Processing Agreements available for signature.
  • Infrastructure Walkthroughs: Our engineering team can walk yours through our security stack.
  • Track Record: Our existing enterprise customers have successfully cleared us through their own rigorous security reviews.

Data Residency (EU & US)

Our pipelines are localized to ensure data sovereignty.

🇪🇺 EU Pipeline (Standard)

Data ingestion, reasoning, and storage occur entirely within the EU region.

🇺🇸 US Pipeline

A US-only option is available upon request.

GDPR Compliance

Parse is fully compliant with GDPR. We never sell, share, or repurpose your data. It is used strictly to provide insights to you, never for marketing purposes.

🧠 3. The Agent Architecture

Concept: Parse is not a "Chatbot Wrapper." It is a deterministic system built on a Multi-Node LangGraph.

Deterministic Processing vs. LLMs

We separate "Data Processing" from "Insight Generation" to ensure security and accuracy.

Raw Data Isolation

Your raw commercial data (transactions, PII, logs) is not sent to third-party LLMs. We use deterministic agentic workflows to process, correlate, and calculate data locally within our secure infrastructure.

AI Usage

We use LLMs only for natural language generation on aggregated, non-sensitive outputs (e.g., summarizing a trend line).

No Training

Your data is never used to train or fine-tune our models. Our reasoning engine is stateless and purpose-built.

How Reasoning Works (The 4-Step Process)

1. Ingest

Fetch raw commercial data via read-only APIs.

Run Python-based math (not LLM math) to ensure 100% numerical accuracy.

3. Reason

Apply deterministic rules to cluster patterns (e.g., "High Churn" + "Open Bug Ticket").

4. Synthesize

Use AI to generate the final human-readable summary from the safe, aggregated data.

🔌 4. Integrations (Technical)

Connecting Event Streams (PostHog)

Type: S3 Export (Batch)

To correlate product usage with revenue, we recommend connecting PostHog via Batch Export.

  1. 1

    Navigate: PostHog → Settings → Project API Keys.

  2. 2

    Create Key: Name it "Parse".

  3. 3

    Scopes: Enable project:read and batch_export:write.

Why Write Access? We require batch_export:write solely to configure the export destination to Parse's immutable S3 bucket. We do not write to your event stream.

Connecting Analytics (Mixpanel)

Type: Service Account

We use Service Accounts to maintain a stable, non-human connection to the Mixpanel Export API.

  1. 1

    Navigate: Settings → Project Settings → Service Accounts.

  2. 2

    Create: Add new Service Account. Role: Analyst (Read-only).

  3. 3

    Input: Provide the Username and Secret in the Parse dashboard.

Database Connections (Postgres / BigQuery)

Parse supports direct connections to warehouses for deeper analysis.

Network

Whitelist Parse IPs: 34.x.x.x, 35.x.x.x (See dashboard for full list).

User

Create a parse_read_only user with SELECT permissions on specific schemas.

⚙️ 5. Custom Agents

Using the Natural Language Builder

You can expand beyond the default library using natural language.

Example Input:

"Alert me if any Enterprise customer (> $50k ARR) has an open support ticket older than 48 hours."

Translation:

Parse converts this into a directed graph query across Salesforce (ARR > 50k) and Zendesk (Ticket Age > 48h).

Questions? Contact us at hello@getparse.io