Technical FAQs
For developers, CTOs, and technology partners evaluating ReadyIntelligence. This page covers architecture, security, integration patterns, and extensibility. If you're looking for general product questions, see our FAQs. For anything not covered here, get in touch with our team.
Architecture and infrastructure
Where is the platform hosted and how is data isolated between clients?
ReadyIntelligence is a multi-tenant SaaS platform hosted within a choice of US or EU infrastructure. Each client deployment runs in its own isolated environment with a dedicated database and storage — client data cannot cross tenant boundaries. The AI infrastructure uses tenant-level namespacing to ensure queries are always scoped to the correct organisation. This is enforced at the infrastructure level, not just application configuration.
Is the vector search handled by a third-party cloud service?
No. ReadyIntelligence operates its own self-hosted vector search infrastructure. Documents are indexed and stored internally — content is not sent to a third-party embedding API for processing. This is a deliberate architectural decision to keep your organisation's knowledge within controlled, permissioned infrastructure.
Does the platform support bringing your own LLM credentials?
Yes. Clients can supply their own LLM API credentials rather than relying solely on the platform's shared model access. This is useful for organisations with existing enterprise AI agreements, or where transparency over compute costs is a requirement. The platform abstracts model selection, so changing credentials or switching models doesn't require application-level changes.
Is ReadyIntelligence tied to a specific AI model or provider?
No. The platform is model-agnostic by design. It uses leading large language models and can be configured to work with different providers. Model selection is handled at the infrastructure level, not the application level — which means agent behaviour stays consistent even as underlying models are updated or changed.
Security and data access
How does ReadyIntelligence prevent the AI from accessing data it shouldn't?
The platform uses a default-deny data access model. By default, the AI has access to a minimal, explicitly defined set of data. Any additional data objects must be registered and permissioned by a developer or administrator before the AI can query them. Permissions are checked at query time against the authenticated user's access rights — so a user can only receive information they are already authorised to see. This is enforced in code, not toggled via configuration.
What authentication and SSO options are available?
ReadyIntelligence supports third-party authentication through identity providers you register, so users of your own site are recognised by the embedded assistant and get permissions and personalisation that match your organisation.
Two methods are available:
- OpenID Connect (OIDC) — users sign in through your identity provider in the browser. You supply a discovery document URL, client ID and client secret, and register ReadyIntelligence as an application with your IdP using the redirect URI shown in the admin. Standard OIDC-compatible providers, including Entra ID (Azure AD) and Okta, can be integrated this way.
- JWT — your host page supplies a signed token representing the current user. ReadyIntelligence verifies the issuer, audience and signature against the key and algorithm you configure.
Identity providers can carry custom claims (for example groups or membership tier), which can drive access rules, be mapped to clearer labels in the admin, and be forwarded as HTTP headers to remote MCP servers using placeholders such as ${user.email}.
What are the platform's compliance certifications?
Pixl8 Group holds ISO27001:2022 and SOC 2 Type II - find out more here.
Extensibility and developer tools
Can developers define custom AI agents?
Yes. ReadyIntelligence provides a handler-based convention for developers to define system agents in code. Custom agents can specify their own instructions, user-facing prompt logic, response processing, and model size preference. Agents are called via a simple service call — infrastructure concerns such as model selection and credential management are handled by the platform, so agent code stays decoupled from the underlying stack.
Can the AI query custom application data, not just the knowledge base?
Yes. This is one of the platform's core differentiators. Developers can register application data objects — including CRM records, membership data, and custom system objects — making them available for the AI to query directly. Each object registration specifies which operations are permitted and what permission rules apply. This allows the AI to go beyond document retrieval and answer questions grounded in live application data — for example, returning a member's current renewal status from your CRM rather than a generic answer from a help article.
Can the AI be triggered by external events or on a schedule?
Yes. ReadyIntelligence supports both webhook-triggered agents (HTTP-triggered background agent runs with logged history) and scheduled agent calls (time-based triggers with run history). This enables automated workflows — for example, a nightly agent that flags at-risk members based on CRM data, or a webhook-triggered agent that routes and summarises a submitted form. These patterns are available to developers without bespoke integration code.
What is MCP and how does ReadyIntelligence support it?
MCP (Model Context Protocol) is an open standard for connecting tools and data sources to AI systems. ReadyIntelligence supports MCP in two directions. First, external services can be connected to ReadyIntelligence via MCP endpoints without custom development work. Second, ReadyIntelligence can itself be exposed as an MCP endpoint, meaning third-party AI clients can call ReadyIntelligence tools and data directly — with full server-side permissioning enforced. This is relevant for partner scenarios where an organisation's knowledge and data needs to be surfaced inside other AI tools.
Integration and deployment
Can ReadyIntelligence be embedded into non-ReadyMembership sites?
Yes. ReadyIntelligence provides an embeddable widget that can be dropped into any website via a standard HTML embed pattern. The full AI capability — including knowledge base search, tool calls, and permissioned data queries — is available on any front-end, including WordPress, Kentico, and third-party member portals. This is actively used by technology partners to surface ReadyIntelligence assistants within their own products.
Is there an API for pushing content into the knowledge base?
Yes. ReadyIntelligence supports data ingestion via API, enabling partners and custom integrations to push content into the knowledge base programmatically. Combined with scheduled web crawls and MCP endpoint ingestion, this gives organisations multiple pathways for keeping content current — useful for mobile app partners or any integration where content originates outside the core platform.
How are AI usage and compute costs tracked?
Token usage is logged at the conversation message level, capturing both token counts and associated costs. This enables reporting on AI usage by client, assistant type, or time period, and supports fair usage modelling. Clients using their own LLM credentials have direct visibility into their own compute costs via their chosen provider.
Working with ReadyIntelligence as a partner?
If you're a technology partner, systems integrator, or developer evaluating ReadyIntelligence for a client deployment, our team can provide more detailed technical documentation, sandbox access, and onboarding support. Find out more about partnerships or talk to our team.