Model Context Protocol (MCP) Services

When an internal AI assistant draws on regulatory data, your organisation may be asked to show where that data came from. Release.art provides Model Context Protocol (MCP) services that allow organisations to connect internal AI assistants and automated workflows to authoritative UK regulatory and corporate data in a controlled, auditable way.

Built by Release.art and delivered through Orlovs Consulting engagements.

MCP services can support multiple internal systems and workflows.

What MCP provides

Our MCP servers offer a structured, read-only interface for querying trusted external data sources.

Unlike data ingestion or embedding approaches, MCP allows AI systems to access current, authoritative information without copying, scraping, or incorporating that data into model weights.

This enables organisations to:

Ground AI systems in authoritative regulatory and corporate data
Maintain control over how and when data is accessed
Preserve clear data provenance and attribution
Support audit, risk, and compliance review
Avoid uncontrolled replication of sensitive or regulated information
Query the source at the time of use, rather than relying on a stored copy

Available MCP data services

FCA Regulatory Data MCP

This service provides controlled, read-only MCP access to the Financial Conduct Authority (FCA) Financial Services Register.

It is commonly used to support:

Internal compliance and risk tooling
Regulatory status and permissions lookups
AI-assisted research and reference workflows
Systems that require authoritative regulatory context

The service is intended for internal systems and workflows only.

UK Companies House MCP

This service provides read-only MCP access to data published by Companies House.

Typical internal uses include:

Company identity and status verification
Director and ownership context for internal processes
Due diligence and onboarding support
Internal research and reporting tools

Access is structured and designed to support traceability and review.

Both services are built on open-source MCP servers published under the MIT licence: fca-mcp and companies-house-mcp.

Designed for regulated organisations

That means:

Read-only access by design
Clear data provenance and source attribution
Observable and reviewable access patterns
Suitable for internal audit and assurance processes
Designed to integrate with existing security and access controls

All MCP services are intended for internal use within client organisations and are not exposed directly to end users or the public.

MCP does not replace human judgement. It supplies trusted context to systems that support decision-making.

Typical use within client environments

Organisations typically use MCP services as a data foundation for:

Internal AI assistants used by compliance, legal, or operations teams
Decision-support workflows that require authoritative context
Automated processes where correctness and traceability are critical
Systems operating under audit or regulatory scrutiny

Specific workflows and applications are implemented separately and remain under client control.

Procurement and audit summary

Scope and intent

Provides controlled access to external regulatory and corporate data
Supports internal AI-assisted workflows only
Does not make compliance, legal, or regulatory decisions

Data access model

Read-only access by design
Source data is fetched at query time and is not trained or embedded into models; query results do enter the model’s context
Clear attribution of data sources

Governance and control

Designed to integrate with existing access control models
Access patterns can be logged and reviewed
Suitable for internal audit and assurance activities

Risk posture

Reduces risks associated with ad-hoc data ingestion
Improves defensibility through clear provenance
Keeps data access separate from decision-making

Limitations and safeguards

Explicit limitations

MCP services:

Do not make regulatory, legal, or compliance decisions
Do not provide legal or regulatory advice
Do not operate autonomously
Do not guarantee regulatory outcomes

They provide contextual data access only.

Safeguards by design

Read-only access to external data sources
Clear separation between data access and decision-making
Designed for human-in-the-loop systems
Transparent and traceable use of source data

Extending MCP access

Additional MCP data sources can be added on request, subject to:

Data licensing and usage terms
Intended use and access model
Security and governance considerations

If you have a specific regulatory, public, or internal dataset in mind, we can assess suitability and implementation options.