Software vendors round up. We made that structurally difficult: the figures on these pages are counted out of the codebase on every build, and a mismatch stops the deployment. This page exists so you can check our arithmetic instead of trusting our adjectives.
What you would be running
Not a roadmap. Every figure here is counted out of the running code on each build, and a mismatch stops the deployment.
16
Specialist AI agents
10
Stakeholder portals
30
Country configurations
25
Native currencies
8
Supported locales
55
Workflow categories
16
Access roles
39
Live MCP tools
6
Messaging channels
12
Accounting adapters
6
AI providers routed
For technical diligence
File counts say nothing about whether a platform collects your rent, so they come last. They are here because a technical evaluator will ask, and because each one is counted from the repository on every build.
320
Application pages
966
API routes
390
Database migrations
764
Test files in CI
208
Data-layer modules
377
UI components
129
TypeScript type modules
15
Security modules
How the honesty is enforced
A promise to keep marketing truthful decays the moment the team gets busy. A test that turns the build red does not. Terraa's test suite treats a stale customer-facing claim as a defect of the same class as a broken query.
Every count on this website is imported from a single module in the codebase. No page, no locale, and no sales document is allowed its own copy of a number. Change the code, and the claim changes with it — or the build fails.
A test walks the repository and counts the real files — pages, routes, migrations, agents, country configs, adapters — then asserts each one equals what this site claims. If the codebase grows and a figure is not updated, continuous integration goes red and nothing deploys.
Marketing copy spells numbers out — “sixteen agents”, “thirty countries”. Those are opaque strings that no ordinary check can read, and that is exactly where stale claims used to hide. A dedicated test now resolves number words back to integers and compares them against the source of truth, so prose cannot drift either.
A layered set of static-analysis tests reads the source and fails the build if any query, AI agent, analytics module, or reporting path can reach data outside the client it was scoped to. The guarantee is enforced mechanically at every layer rather than reviewed by hand.
Tests fail if a user-facing string is added without a translation key, if a key exists in English but is missing from any other locale, or if a translation is left empty. An Arabic or Urdu operator cannot silently be shown English.
Layout rules — minimum touch-target sizes, mobile-first grids, responsive widths, drawer behaviour, right-to-left correctness — are checked by tests against the component source. A screen that would break on a phone does not reach production.
What we do not claim
A proof page listing only strengths is an advertisement. These are the things a careful procurement team would ask about, answered honestly.
Terraa is engineered to SOC 2 principles and can evidence its controls, but it does not hold a completed Type II attestation. Anyone who tells you a young platform is already certified is either mistaken or selling.
Internal security review, automated scanning, and a layered application security stack are in place today. A commissioned third-party penetration test is scheduled and has not yet been completed.
Bring-your-own storage is live today, and the storage layer is provider-agnostic by design so region selection is a configuration change. A fully self-hosted in-region deployment is scoped per customer and is not something we claim as shipped.
Bring your hardest question