Tools
Read and action endpoints the AI can call — look up a record, run a calculation, or make a change. Each tool has a clear contract and validated inputs.
Custom MCP Servers
Securely connect AI to the systems you already run.
MCP — the Model Context Protocol — is the open standard that lets AI assistants safely read from and act within your systems through well-defined, permissioned tools. CTR/NY designs, builds, and operates the servers that make it work on your real data and workflows.
MCP Server
your-systems → AI assistant
# tools the AI is allowed to use
tool get_overtime_by_department()
tool query_warehouse(scope)
tool post_schedule_change()
resource policy_handbook
✓ authenticated · ✓ scoped · ✓ audited
The standard
What is MCP?
The Model Context Protocol is a secure, standardized way to give an AI assistant a defined set of tools and data sources. Instead of handing the AI broad, risky access to everything, you expose only the specific, permissioned actions you choose — each one validated, scoped to the right user, and logged.
In plain terms: an MCP server is the secure bridge between an AI assistant and your business systems. The AI can ask the server to look something up or take a defined action — like "how much overtime did Warehouse run last week?" — but it can never reach past the tools you've published. You decide what's possible. The protocol enforces it.
What we build
The building blocks of a custom MCP server
Every server we deliver is composed of the same well-governed parts — designed around your systems, your roles, and your rules.
Resources
Documents and data sources the AI can reference — handbooks, policies, knowledge bases, and live records — surfaced safely and in context.
Secure connectors
Hardened connections to your databases, APIs, and platforms — including legacy and in-house systems — without exposing credentials to the AI.
Authentication & identity
Every request is tied to a verified user or service identity. No anonymous access — the server always knows who is asking.
Scope & permission enforcement
Results are filtered to what each user and role is allowed to see — row-level scope, enforced on our side, every single time.
Audit logging & monitoring
Every tool call is recorded — who asked, what they requested, and what came back — so you have a complete, reviewable trail.
Don't want to start from scratch? Start from the library.
Our Perigee Cloud connector library ships ready-made, secure connectors for Attendance on Demand, Dynamics 365 Business Central, Paylocity, GT Clocks, Zendesk Support, and more — built on the same MCP foundation. And if your team already uses Claude, a secure MCP connector can link the AI you already pay for to your business data.
Secure by design
Powerful AI access, on a short leash.
Connecting AI to real systems only works if it's safe. Our MCP servers are built around least privilege from the first line of code — the AI gets exactly what it needs and nothing more.
- Least-privilege access — only the tools you publish
- Per-user and per-role scope on every result
- A full audit trail of every request and response
- Your data stays yours — no broad database dumps
Request, enforced
a single tool call, governed end to end
# manager asks the AI a question
query_warehouse(user="j.flanagan")
→ identity verified
→ scope applied: Warehouse only
→ result returned · request logged
✓ authenticated · ✓ scoped · ✓ audited
End to end
Designed, built, and maintained by CTR/NY.
An MCP server is not a one-time install — it's a living connection to systems that change. We own the full lifecycle so your team doesn't have to, backed by 75+ years of getting workforce data right.
- Architecture — we map your systems and design the right tools
- Security review — least-privilege scopes, identity, and audit
- Deployment — wired into your environment and AI assistant
- Ongoing maintenance — monitoring, updates, and new tools
From design to operations
one accountable partner
01 map_systems() # discovery
02 define_tools() # architecture
03 review_security() # scope + audit
04 deploy() # into your env
05 monitor() # ongoing
✓ designed · ✓ secured · ✓ maintained
Give your AI safe access to your real systems.
Let's scope a custom MCP server around your data and workflows — and show you exactly what your AI could do, securely, in a live demo.