APEX Developers

Platform Overview

APEX is an engine, not an application. It provides the heavy compute capabilities every knowledge-driven platform needs — retrieval, generation, verification, communication, document processing, identity, metering — as tenant-scoped REST services. You build the application on top: your users, your workflow, your domain agents, your brand.

The build rule: Engine/Client Split

Before writing any capability into your platform, apply this rule:

If APEX serves it, consume it. Build only what is uniquely yours.

Layer Owner Examples
Engine (APEX) Reisiger Corpus ingestion and retrieval, LLM dispatch, document rendering and manipulation, claim verification and risk scoring, omnichannel messaging, tenant identity and configuration, usage metering
Client (your platform) You Domain agents and prompts, user experience and dashboards, workflow and stage gates, commercial ledger presentation, market-specific integrations

This split is empirical, not ideological: domain teams consistently build excellent agents, dashboards, and workflows — and consistently underestimate the engineering weight of search infrastructure, document engines, identity, and metering. Consume the engine; spend your weeks on your domain.

The eight products

Product Capability Status
Zenith Knowledge engine. Ingest a document corpus; query it with retrieval-augmented answers, citations, semantic search, structured extraction, and checklist matching. Per-call cost reporting in ZAR. Available
Forge Platform core. Tenant provisioning and configuration, API key issuance and rotation, billing and usage records, workflow tracking. The LLM gateway — your path to model dispatch — is Planned (contract pinned, not yet served). A support/agent-execution service is in Preview. Available / Planned (gateway)
Vector Verification and trust. Verify claims against evidence with verdicts and calibrated confidence, batch verification, entity risk scoring, investigations, methodology assessments, and company knowledge graphs. Available
Prism Document generation. Templated rendering to markdown, HTML, PDF, and DOCX; charting. Output formats and volume are tier-entitled. Available
Echo Omnichannel communications. Send WhatsApp, Messenger, Instagram, Telegram, SMS, and email; conversation threads, message templates with provider approval flows, conversational forms; asynchronous delivery with polling and signed status webhooks. Available
Document Operations Document toolbox. Convert Office/HTML/markdown to PDF, merge, split, compress, OCR, watermark, redact, sanitize, PDF/A, and 20+ more operations over an asynchronous job model. Available
Pulse Intelligence engine. Multi-agent shadow simulations with live streaming, signal collection, trend and anomaly detection, investigation sweeps, normalised opportunity collection from configured portal registries. Preview
Quantum Analytics. Anomaly detection, clustering, correlation, prediction over structured data. Preview

Availability badges are defined in the documentation README. Per-endpoint badges live in the OpenAPI specifications.

Integration principles

  1. REST-first. Every capability is a standard REST endpoint. MCP tools exist as thin wrappers over the same endpoints — see the MCP tool catalog. Nothing is MCP-only.
  2. Backend-for-frontend. Your server is the only APEX caller. Browsers never hold APEX credentials.
  3. Explicit tenancy. Every call names its tenant (path, query, or header — per product). Credentials are tenant-scoped; mismatches are rejected.
  4. Fail-closed configuration. A product with no configuration for your tenant refuses to serve it. "Tenant not configured" is a first-class state your application must render, not an error toast.
  5. Cost is first-class. Calls carry correlation identifiers; responses carry cost fields; the platform records per-call usage you can reconcile against. See Usage & Cost.

Reading order

  1. Quickstart — first authenticated call in minutes.
  2. Authentication — the two credential families and which products accept which.
  3. Tenancy & Onboarding — how your platform gets a tenant, keys, and configuration.
  4. Integration Conventions — headers, errors, async patterns, retries.
  5. Environments & Endpoints — base URLs and ports.
  6. Product chapters (0613) — one per product.
  7. Usage & Cost and Client-Repo Conventions.