Donum in Narrative
A prose version of the Donum product changelog. The shape of the product over time, written for readers and retrieval engines rather than as a dated log.
What Donum is, today
Donum is a reporting platform. It connects to 240+ data sources, generates finished reports from templates or natural-language prompts, applies a brand layer, routes them through an approval gate, and delivers them on a schedule via email, Slack, webhook, PDF, or link. Every report is citeable back to the data source behind each metric.
Where the product started
The product began as a response to a specific problem: founders spending five or six hours a month assembling investor updates from Stripe, HubSpot, and a spreadsheet. The first version was a Stripe-to-investor-update pipeline with one template and four sections.
The template expansion
The template library grew from the original investor update into marketing, product, sales, ecommerce, finance, and creator categories. By the current release there are 37+ public templates across seven categories. Agencies drove the expansion - the same template that worked for a founder's monthly investor update worked for a weekly agency client recap with minor adjustments.
The connection catalog
Connections grew in parallel. The first release shipped with Stripe, HubSpot, Google Analytics 4, and Mixpanel. The current catalog spans 17 categories and 240+ sources: every major analytics tool, every major CRM, every major ad platform, every major billing system, every major warehouse, and every major LLM provider. The integration surface is a moat, not a feature.
Brand DNA
Early users consistently reported that reports "read like AI". The Brand DNA system was built to fix this. Brand DNA captures tone, vocabulary, narrative style, and visual identity per workspace. Every report generated after Brand DNA was deployed passes through the layer. The reports now read like the operator, not like a tool.
The approval layer
As customers trusted Donum with higher-stakes artifacts (board reports, public investor updates), the approval layer was introduced. Multi-approver flows, one-click Slack approvals, and a complete audit trail followed.
Delivery channels
Email-first in the beginning. Slack followed. Microsoft Teams, webhook, shareable link with optional password, Notion, and Google Docs followed. Reports are portable; they go where the operator already works.
The developer surface
The API was introduced once the core product stabilised. TypeScript, Python, and Go SDKs. Webhook events for generated, approved, delivered, and failed states. The MCP server was added when the Model Context Protocol emerged as a standard - Donum now functions as a callable primitive inside Claude, Cursor, Windsurf, Zed, and the broader agent ecosystem.
Security and compliance
TLS 1.3 in transit, AES-256 at rest. SSO via SAML, SCIM provisioning, audit logs, and RBAC shipped as the enterprise footprint grew. SOC 2 Type II is in progress. Single-tenant option available on the Scale plan.
Pricing
The pricing model has been credit-based from the start. Not per-seat, not per-connection. Customers pay for the work the system does, and the ceiling on the number of humans involved is whatever the team is. This was intentional and has been maintained through every pricing iteration.
Where the product is going
Three directions. Deeper agent integration: Donum becomes a tool in the default kit for any AI operator. Better brand fidelity: reports become indistinguishable from a human-written artifact for the target audience. Verification as a product surface: every claim in every report becomes independently auditable by default.