Claude Dashboards & MCP

Put Claude to Work on your Telemetry

We build MCP servers that give Claude controlled, logged access to your telemetry and systems, and Claude dashboards your team can question in plain language. Read-only by default, widened only with your approval.

Our Claude services

Two Services That Work Together

An MCP server is how Claude reaches your data safely. A Claude dashboard is how your team gets answers from it. You can start with either, and most teams end up with both.

The connection

MCP Servers

Custom Model Context Protocol servers that give Claude scoped, logged access to your telemetry, Grafana and Spotfire data, and internal APIs.

  • Tools, inputs and limits agreed with you before any code
  • Read-only first; anything that writes waits for a person
  • OAuth or token auth on your existing roles
  • Every call logged, every tool tested against misuse
Explore MCP Servers
The answers

Claude Dashboards

Dashboards your team questions in plain language, answered from live, governed data, with the query behind every number shown alongside it.

  • Ask a question, get a chart and the query that made it
  • Daily and weekly briefs on what changed and why
  • Each person sees only what their role allows
  • Runs alongside the Grafana and Spotfire boards you already use
Explore Claude Dashboards

How it fits together

From Your Data to a Checked Answer

  1. 01

    Your data

    Prometheus, InfluxDB, Grafana data sources, Spotfire tables and internal APIs, staying where they are.

  2. 02

    MCP server

    Scoped, parameterised tools with your authentication and a log of every call.

  3. 03

    Claude

    Picks the right tools for the question and treats what comes back as data, never as instructions.

  4. 04

    Claude dashboard

    A chart and a plain-language answer, with the source query so anyone can check it.

What a Claude dashboard looks like

Ask the Question Nobody Charted

Grafana and Spotfire show what someone decided to chart in advance. A Claude dashboard answers the question you have right now, from the same governed data, and shows its working.

  • “Which sites breached their latency target this week, and since when?”
  • “What changed before last night’s spike in packet loss?”
  • “Summarise yesterday’s alerts for the morning stand-up.”
  • “Compare throughput at our five busiest sites against last month.”
Claude dashboard · Network ops

What changed before last night’s spike in packet loss at Site 14?

Packet loss at Site 14 rose from 0.2% to 3.8% between 01:10 and 01:40. It started four minutes after a config push to edge router r14-edge-2, and returned to normal when that change was rolled back at 01:38.

config push 01:06 00:30 02:30
Source query avg_over_time(packet_loss_ratio{site="14"}[5m])

Read-only · ran as ops-viewer · call logged

Illustrative example

Everything included

What We Build

When you’re ready

See Claude working on your own data.

Claude & MCP In Depth

Useful AI That Earns Its Access

What do you build?

Two things that work together: custom MCP servers that connect Claude to your telemetry, Grafana, Spotfire data and internal systems, and Claude dashboards that let your team ask questions of that live data.

Why MCP instead of pasting data into chat?

Pasted data is a snapshot. It goes stale the moment you copy it and has no controls around it. An MCP server gives Claude live access through a defined set of tools, so answers come from current data and every call can be logged and reviewed.

How is a Claude dashboard different from Grafana or Spotfire?

A Grafana or Spotfire dashboard shows what someone decided to chart in advance, and it is the right tool for what you watch every day. A Claude dashboard can be asked a question nobody planned for and can regenerate a view on demand. We use all three on the same governed data.

What can Claude do with our data?

By default it can read: summarise the last day, compare sites, show a trend, or give context around a spike. It cannot change anything. Actions are only added where you explicitly approve them, one tool at a time.

How do you keep it safe?

Least privilege: each tool exposes only what its job needs. Authentication follows your existing permissions, every call is logged, and data returned to Claude is treated as untrusted input, since a text field in a database can contain instructions. Any write needs human approval.

Can we trust the answers?

Numbers come from queries, not from the model's memory. Each answer shows the tool calls behind it, so an engineer can check the query, the time range and the units. We add checks for the common mistakes, such as a wrong window or mixed units.

How do we start?

With a read-only pilot on one question your team asks often. We build one MCP server with a handful of tools, connect it to Claude, and build one dashboard on it. You judge it by use before anything is widened.

Read-Only by Default

Claude can read your data. Anything that writes needs explicit approval.

Every Call Logged

Each tool call is recorded in an audit log you can review.

Specialist Delivery

Engineers who build MCP servers and data platforms for production

Why Choose Us

AI That Has To Earn Its Access

MCP lets Claude reach your tools and data in a controlled way. We design every server around the minimum its use case needs, and every dashboard around a question someone actually asks.

Data returned to Claude is treated as untrusted input, every call is logged, and anything that changes a system waits for a person to approve it.

Least Privilege

Each tool exposes only what its job needs, nothing more

Answers With Sources

Every number traces back to the query that produced it

Sit down with us

We believe AI should earn its access, one question at a time.

Analyst reviewing a KPI dashboard on a laptop
3

Three access levels: read, scoped, approved.

How we work