Home/Work/Production MCP Servers

Production MCP Servers

I help you connect Claude — or any AI model — to your own systems, so you can query them in plain language. Questions are answered live and scoped to who's asking. All secured by OAuth.

Ask your business data.

Just try it for yourself: this demo calls a live MCP server and answers questions about my portfolio, plus live crypto and FX data.

Claude Claude · connected to mcp.danielsolves.ai
scope public demo · live data, no login
● demo
Live data through MCP. Pick a question below — I'll call the real tool and answer from the source.
Markets
Ask Daniel

Plug this server into your Claude.

This demo server is open and read-only — public data only. Point your own Claude at it and ask the same questions in your own client.

Claude Desktop & Claude.ai

Settings → Connectors → Add custom connector → paste the URL:

https://mcp.danielsolves.ai/mcp

Claude Code

One command in your terminal:

claude mcp add --transport http danielsolves https://mcp.danielsolves.ai/mcp
9 tools available
crypto_price() current price + 24h change
crypto_history() price time series (powers the chart)
top_crypto() top coins by market cap
fx_rate() convert / current FX rate
fx_history() FX time series
daniel_profile() who Daniel is, how he works
daniel_projects() projects, filterable by tech / area
daniel_skills() skills, filterable by area
daniel_availability() availability + how to engage

The data existed. Getting at it didn't.

Every answer the team needed was technically in the system — orders, analytics, customer records. But getting it meant knowing the right dashboard, the right filter, or pinging the one person who could write the query. The AI everyone already used couldn't touch any of it.

Before
  • Simple questions queued behind whoever could run a report
  • Answers scattered across dashboards, exports and tribal knowledge
  • Claude / ChatGPT useful for text — blind to the company's own data
  • No safe, auditable way to let AI read from production systems
After
  • Ask in plain language — the AI calls the right tool and answers
  • One conversational surface over search, analytics and reporting
  • Scoped, OAuth-secured access — every call authenticated and logged
  • New capabilities added as tools, without retraining anything

A clean bridge between the model and your systems.

MCP (Model Context Protocol) is the open standard for giving an AI safe, structured access to tools. I build the server that sits in the middle — exposing your data as well-defined, permissioned tools the model can call.

01 · AI client

Claude asks

The user types a question. Claude decides which tool can answer it and calls it with structured arguments.

→
02 · MCP server

The bridge

My server authenticates the request via OAuth, runs the tool against your real systems, and returns structured data.

→
03 · Your data

Source of truth

Databases, internal APIs and reporting — queried live, scoped to exactly what that user is allowed to see.

Tools, tailored to the domain.

Each MCP server exposes a focused set of tools. The model only ever sees what it's allowed to call.

crypto_price()

Live prices

Current crypto price and 24h move, pulled on demand from a cached public feed — never stale.

crypto_history()

Time series

Historical price points the model can summarise — and the data behind the inline chart in the demo.

top_crypto()

Rankings

Top coins by market cap, sorted and trimmed to the count you ask for.

fx_rate()

FX & convert

Live exchange rates and currency conversion against an ECB reference — plus fx_history() for the trend.

daniel_projects()

Structured profile

Projects, skills, profile and availability as queryable data — a CV the model can actually read.

+ custom

Your domain

Point the same pattern at your systems — each new capability is a new tool, scoped to who's asking.

Built for production, not a demo.

OAuth on every call

Each tool invocation is authenticated. No anonymous access to production data, ever.

Scoped permissions

The model only sees tools — and rows — a given user is allowed to access. Least privilege by default.

Full audit trail

Every call is logged with who, what and when — so AI access is as accountable as any other system.

Typed contracts

Strict input/output schemas mean the model can't send malformed calls or get surprising data back.

Tech stack.

Claude APIMCP ProtocolTypeScriptNestJSOAuth 2.0PostgreSQLREST / internal APIsDocker

Want your team to just ask your data?

If your answers are locked behind dashboards and queries, MCP can put them one sentence away — safely. Tell me about your systems; I'll reply within 2 hours.

Start a project