Ask Claude about your data
Read-only access to your Postgres, Supabase or analytics database, so your team can ask questions in plain English instead of waiting for someone to write SQL.
Back to the homepage01 / The idea 02 / What I can build01 02 03 04 05 06 03 / Packages 04 / Safe by default 05 / How we work01 02 03 04 06 / Something I’ve built 07 / Questions
Services / MCP & Claude Code
Claude, wired intothe way your team works.
I build custom MCP servers and Claude Code workflows, so your team can ask Claude to work with your own database, APIs and tools, safely.
- Starting at
- $500
- Typical delivery
- 1–2 weeks
- Works with
- Claude Code, Claude Desktop, any MCP client
- Built with
- Python or TypeScript
The idea
Your team already uses Claude. But it can’t see your database, your internal tools or your deploys, so people copy and paste between windows. An MCP server gives Claude its own carefully limited access, so it can do the work itself.
A few ideas to start with.
Your internal API, as tools
Look up a customer, check an order, create a ticket. Claude calls your own API with the same permissions a person would have.
Deploys and operations
Check deploy status, read recent logs and run the safe, pre-approved actions your team does every day.
Claude Code for your team
Shared commands, hooks and review checklists that teach Claude Code your repo’s conventions, so everyone gets the same good results.
Other machines and environments
Run Claude Code work on a build machine, a server or a second Mac, like my open-source claude-bridge.
Your docs and knowledge
Let Claude search your documentation, wiki or support history and answer with links to the source.
Clear scope. Fixed prices.
Single MCP server
From $500One system connected to Claude: your database, an API or an internal tool.
- A focused set of tools, scoped together with you
- Authentication and least-privilege access
- Tests and setup instructions
- A handover call so your team can run it
Team workflow
From $1,500Claude working across several of your systems, set up for a whole team.
- Several connected tools or MCP servers
- Claude Code commands and hooks for your conventions
- Rollout to your team, with docs
- Two weeks of fixes after launch
Something bigger or different? Tell me about it and I’ll send a written scope.
Access you can explain to your security team.
An AI with access to your systems has to be built more carefully than a normal integration. Every project starts from these rules:
- Read-only first; write access only for the actions you choose
- Each tool gets the smallest permissions it needs
- Secrets stay on the server, never in prompts
- Risky actions require a person to approve them
From a call to a working tool.
A short call
You tell me which systems you want Claude to use and what your team does with them today.
A written scope and a fixed price
Exactly which tools I’ll build, what they can and can’t do, and what it costs. No surprises.
Build and review
You see it working against a test environment before anything touches production.
Handover
Your code, in your repository, with docs. You own all of it.
claude-bridge.
An open-source MCP server that lets a Claude Code session run real, resumable tasks on another Mac over a private Tailscale network, with token auth, allowed folders and careful permission defaults.
Good to know.
What is MCP?
The Model Context Protocol is an open standard for giving AI assistants tools. An MCP server is a small program that tells Claude what it can do, such as “look up an order”, and does it safely when asked.
Do we need Claude Code?
No. MCP servers work in Claude Code, Claude Desktop and other MCP clients. Claude Code workflows need Claude Code, of course.
Is it safe to give Claude access to our systems?
Only if it’s built carefully. I start read-only, give each tool the smallest access it needs, keep secrets on the server and make risky actions require approval.
Who owns the code?
You do. It’s delivered to your repository, and you can change, extend or host it however you like.
Can you work with our existing stack?
Usually, yes. If your system has an API, a database or a command-line tool, it can almost always be connected.