TRIRIGA and MREF toolsbyAJAEX

Deployment and security

Who runs what, and where your data goes.

Start with your approved AI client and AJAEX's hosted tools. Compare who runs each component, where data travels, and the hosted and local options.

Common architecture

The bridge stays with TRIRIGA. The intelligence layer can move.

Layer 1

Intelligence and control

Semantic routing, evidence composition, governed MCP and REST tools, tenant and environment policy, confirmation, and audit.

AJAEX-operated for Bring your own AI, with or without our AI model. Customer-operated for the local estate licence.

Layer 2

Customer-side bridge

Supported in-JVM access to the TRIRIGA platform surfaces that require local execution, licensed for the named scope.

Installed and operated with the customer in every model.

System of record

Customer TRIRIGA

Platform, database, configuration, records, documents, runtime, and the approved integration identity remain customer-hosted.

Ways to run it

Compare responsibility, data flow, and current availability.

Available now, recommended

Use our AI model

Most teams: our workspace and AI model on your own TRIRIGA instances, with a usage allowance.

AJAEX: The workspace, model, hosted tools, gateway, tool policy, audit log and administration.

Customer: TRIRIGA, the installed bridge, and the named users.

Available now

Bring your own AI

Teams that want to pay for inference on their own model account (Anthropic, OpenAI or AWS Bedrock), under their own contract.

AJAEX: The workspace, hosted tools, gateway, tool policy, audit log and administration, connected to your inference provider.

Customer: The model account, TRIRIGA, and the installed bridge.

Priced on your estate

Local estate licence

Estates that want the tools in their own cloud account or premises, on their own model.

AJAEX: Releases, set-up help and support for the licensed distribution.

Customer: The host, AI model, secrets, logs, network, TRIRIGA and operating controls.

Responsibility matrix

Who operates each component.

Scroll horizontally to compare all models.

Responsibility by deployment model
ComponentBring your own AIUse our AI modelLocal estate licence
UI and user experienceCustomerAJAEXCustomer
AI model or model accountCustomerAJAEX and its model providerCustomer
Intelligence, MCP, and admin layerAJAEXAJAEXCustomer
Bridge inside TRIRIGACustomerCustomerCustomer
TRIRIGA and system of recordCustomerCustomerCustomer

Data flow

Bring your own AI keeps the AI client with you; tool data still passes through AJAEX's gateway.

Each service design documents regions, provider terms, retention, backup, support access, deletion, and subprocessors.

Scroll horizontally to compare all models.

Data flow by deployment model
Data categoryBring your own AIUse our AI modelLocal estate licence
Prompts and conversationsYour AI client and model provider; AJAEX only if your client sends themAJAEX workspace and its model providerCustomer systems and selected provider
Tool arguments and resultsAJAEX gatewayAJAEX gatewayCustomer-hosted tooling
Service credentialsAJAEX secret store for the named environmentAJAEX secret store for the named environmentCustomer secret store
Identity and audit metadataYour AI client plus AJAEX gateway auditAJAEX workspace, admin, and gatewayCustomer-hosted audit
TRIRIGA system of recordCustomerCustomerCustomer
Licence, update, and support trafficDocumented for the selected tooling serviceDocumented for the hosted workspaceOnline, offline, or proxied route agreed during solution design

The gateway runs in AWS London (eu-west-2). Any contractual commitments for residency, retention, backup, support access, or recovery are set out in the selected service agreement.

Control questions

Review the controls behind each deployment model.

AJAEX LTD holds Cyber Essentials Plus (renewal 13 November 2026). Cyber Assurance and ISO 27001 certification are in progress. It holds no ISO 27001 or SOC 2 certificate today, and no independent penetration test has been run. The security pack lists each control and each open item.

Service identity and least privilege

The customer supplies a dedicated TRIRIGA integration identity. Its permissions and the enabled tool list together define the maximum reachable scope.

Credentials kept apart

The TRIRIGA service user's password is held in a per-customer secret store. The solution design records who can reach it and how it is rotated.

Tool policy and confirmation

Read, write, administrative, and deployment capabilities are separated. You choose the exact tools for each environment. Write tools stay off until the policy turns them on, and a person confirms each change before it is written.

Explicit data flow

Prompts, arguments, results, credentials, identity metadata, logs, diagnostics, and provider processing are mapped for the selected model.

Customer-operated bridge

The supported bridge is installed inside the customer's TRIRIGA or MREF environment, in the javax or Jakarta build that matches the platform.

Assurance follows the model

Every tool call is logged with environment, action and outcome, without secrets, and kept for 90 days. Ask for the security pack to review retention, backup, support access, recovery and the open items before agreeing the service scope.

Hosted MCP connections authenticate with a credential for your tenant, not one per person. Calls run in TRIRIGA under the tenant's integration identity: per-user TRIRIGA identity and user-level OAuth are not standard controls, so any production engagement that requires them must address that before use.

Make the deployment boundary the first architecture decision.

The architecture review maps the AI, tooling, bridge, identity, network, data flow, enabled tools, and assurance evidence before a guided start begins.