Built around the same public information your team can review on the website.
Prepared examples
Get a useful result before you contact anyone.
These examples use the same deterministic assessment and calculation logic as the MCP tools. They are prepared page content and make no remote MCP calls.
Question
We have 40 employees and spend three days every month consolidating reports from several systems. Where should we start?
Tool usedassess_workflow
Readable result
Start with workflow automation or a focused internal tool
The main value sits in reducing repeated admin, inconsistent handovers and work that depends on individual memory. First version: Centralise one high-volume process with required inputs, routing or approvals, clear status and exception handling. Do not automate every exception before the normal workflow and ownership rules are clear.
What might a workflow automation be worth under conservative assumptions?
Tool usedcalculate_workflow_roi
Readable result
61.9% first-year ROI · 6.8 month payback
Using 800 cases per month, 12 minutes saved per case and a 60% capture factor, the hypothetical result is 1920 annual hours released and R417 600 realised annual benefit. These are assumptions to challenge, not a forecast.
We want an AI assistant to answer recurring questions from operating documents. Are we ready?
Tool usedassess_ai_agent_readiness
Readable result
Start with document intake and review support
The clearest first use case is extracting, summarising and routing information from repeated document workflows. The use case is promising, but the information source is not reliable enough yet. First organise an approved source of truth, define ownership and collect representative requests. Then pilot with draft for human review. Do not ask an agent to compensate for missing, contradictory or unowned business information.
The examples remain available as page sections if JavaScript is disabled.
Connection guide
Three steps to ask a better question.
The exact labels vary between clients, but the flow is the same: add the server, check the tools and ask for a source-backed answer.
01
Open remote MCP settings
Find the area where your assistant manages remote MCP servers or connected tools.
02
Add the endpoint
Use the public endpoint in the connection panel. No API key is required for this catalogue.
03
Ask and inspect
Confirm the eight tools are visible, then ask a business question and review the assumptions, limitations and source links.
Transport Streamable HTTPAuthentication None requiredData boundary Selected public catalogue only
Tool reference
Eight focused ways to plan and explore.
Each tool is read-only. They return structured data and readable content, with practical boundaries so an assistant can choose a sensible next question.
Assess and calculate
01Assess a business workflowassess_workflow+
Input
Problem plus optional audience, current tools and desired outcome
Returns
Practical starting point, first version, constraints and inputs to clarify.
Boundary
Planning guidance only; consider process changes and existing software before custom development.
02Assess AI agent readinessassess_ai_agent_readiness+
Input
Task plus optional audience, reliable source and action boundary
Returns
Readiness signal, prerequisites and a bounded pilot recommendation.
Boundary
Missing or unowned sources return foundation-first guidance.
03Calculate workflow ROIcalculate_workflow_roi+
Input
Seven non-negative numeric assumptions; capture factor is 0–100
Returns
Hours, capacity value, realised benefit, first-year cost, ROI and payback.
Boundary
Hypothetical calculation; released capacity is not guaranteed cash saving.
Explore LCR and prepare a brief
04Search public knowledgesearch+
Input
A short question or topic, plus optional record type
Returns
Matching public records and source links.
Boundary
Use fetch when you need a complete citation-ready record.
05Read one complete recordfetch+
Input
A catalogue record ID
Returns
The public record with id, title, text, URL and selected tags.
Boundary
Unknown IDs return a clear not-found response. Confidential details are never included.
06Match a business problemmatch_service+
Input
Problem plus optional industry and systems
Returns
Relevant capabilities, matched public evidence and discovery questions.
Boundary
A capability match is not a quote, recommendation or acceptance of work.
07Find published workfind_relevant_case_studies+
Input
Problem plus optional technology or industry
Returns
Relevant public work summaries and an evidence boundary.
Boundary
Unsupported technology or client requirements return no published evidence.
08Prepare project questionsqualify_project+
Input
Project context plus optional process, systems, users, outcome, budget or timeline
Returns
Potential fit, missing information and a copyable brief.
Boundary
It offers a website link; it never submits a lead or generates a quote.
Why connect?
A website answers these questions for people. MCP makes them available to assistants too.
Before someone contacts a business, they want to know whether it can help, whether there is relevant evidence and what to do next.
01
Can LCR help?
Service matching connects a described problem to public capabilities and clarifying questions.
02
Has LCR done something relevant?
Published work summaries provide evidence, with a clear boundary around what has not been published.
03
What should happen next?
Qualification surfaces missing information and offers a reviewed contact-form handoff.
MCP
Remote assistants can call structured tools after a client is configured with the endpoint.
WebMCP
Compatible browser agents can discover tools exposed by the page while visiting the site.
Questions
A few useful boundaries.
The interface describes what LCR makes public today. More tools can be added to the same connection over time.
What do I need to connect?+
A compatible AI assistant or MCP client that supports remote Streamable HTTP servers. Add the endpoint below in that client’s MCP settings, then ask a question in the client.
Does the server need authentication?+
No authentication is required for the current public catalogue. The endpoint is read-only and returns selected public information from LCR’s website.
Can it submit an enquiry or give a price?+
No. Qualification can prepare a brief and offer a contact link for someone to review. An enquiry requires a person to review and submit the form, and LCR still needs to discuss scope before pricing.
What is measured?+
LCR records protocol connections and tool outcomes through Boxley using request-scoped IDs, duration, tool name and public record IDs. Raw questions, project descriptions, contact details and complete request or response bodies are not sent. The measurement shows how the interface is used; it does not identify a person or guarantee a lead.
Can I opt out of MCP usage measurement?+
Yes. Connect using the same endpoint with ?analytics=off, or send the X-LCR-MCP-Analytics: off header. This disables Boxley MCP events for that request; it does not change the result.
How long is MCP measurement kept?+
MCP interaction events are retained for up to 90 days for product measurement and troubleshooting. Hosting and security logs may follow the hosting provider’s separate retention controls.
Use it with a real question
Put LCR MCP to work.
Connect an assistant, explore the public catalogue and use the answer to decide whether a focused conversation makes sense.
