About
I'm Chris Bailey.
Bailey Wildash is my one-person consultancy. I get complex integrations live and used, then turn the method into Claude Code skills, instructions an AI agent follows, so your team keeps it after I've gone. I'm based in the UK and work with teams anywhere.
What I do
The stretch between a signed deal and real use
That stretch is where integrations go wrong, and it's rarely anyone's whole job. Sales has moved on to the next deal. Engineering is building what the ticket says. The customer's team is waiting on something from yours. It sits between engineering, product, sales and the customer, and that gap is where I like working. I check what was promised, make sure every dependency has someone's name on it, review the design wherever money or data moves, and test how things really behave rather than assuming.
The other half is agentic work, which in plain terms means AI agents doing real work rather than drafting text. I write the same method down as Claude Code skills and workflows that live with your code, so an agent can prepare for the call, check the docs and draft the plan, and a person can check its work without redoing it. That's what keeps the method going after I've gone.
It's deliberately small. You get senior attention on the actual problem, not a team put together around a statement of work, and a straight answer when something is outside what I do well.
Built the same way
This site is written the same way
A Claude Code skill keeps these pages up to date. It's a small working example of what I build for clients. My working notes are full of detail that must never appear here, so the agent drafts, a plain script checks and I decide.
My notes
It only reads folders I've approved.
What's new
Anything unchanged since last time is skipped.
A draft
Rewritten in general terms, with no clients or specifics.
A privacy check
A script checks the draft against a list of names from my notes, and for personal details, passwords or API keys, and any phrase copied from the notes. One match stops the update.
Found something: the update stops
My review
I read every change.
My sign-off
Nothing goes live until I approve it.
How I work
What you can hold me to
If I break one of these, tell me.
Say where it came from
When I tell you something works, I'll say whether I read it in the docs, saw it in a test or I'm inferring it. A guess doesn't become a fact because it sounds right.
Don't bluff
If I don't know the answer, I'll tell you what I do know, what needs checking and the quickest way to check it.
Automate what comes up twice
When a problem turns up a second time, I write it down, script it or take it to the product team. Handling it by hand every time costs more.
Write it where it lasts
Decisions, blockers and owners go in files your team keeps, not in meeting notes nobody reopens.
Leave things that work without me
Skills, checklists and tests stay with your code and keep working after I've gone.
Say when it isn't worth doing
Sometimes the honest answer is that you don't need me for this one.
Fit
Who I work with
You might have signed a big customer in the spring who still isn't live. You might have gone live on time and found hardly anyone's using it. Or every enterprise deal now comes with an integration project, and your team isn't set up to run them. If your product is an API, a payment flow or a digital asset service, and your customers' engineers have to build against it, that's the shape of problem I'm for.
Size matters less than the shape of the problem. A small developer platform and a large company rolling out a complicated integration can get stuck in exactly the same place.
Boundaries
What I don't do
- Take work outside what I do well
- Publish client names or case studies without permission (why there aren't any here)
- Sell headcount, or resell someone else's
- Recommend a rebuild when a fix will do
- Tell anyone a product makes them compliant. Regulatory interpretation belongs with your legal or compliance owner.
Client work appears on this site only with permission. Nothing's invented in the meantime.
Expertise
Technical ground covered
I start from the problem, not a favourite tool. The one fixed choice is Claude Code for the agent work.
Integration architecture
The calls usually work. What breaks is the same event arriving twice, the second one arriving first, and the timeout where you can't tell whether the payment went through. I design for those across APIs, webhooks, auth, serverless setups and web and mobile SDKs, with one system that has the final say and clear rules for how a status can change.
Testing and evidence
I'd rather prove how something behaves than assume it. I test against the provider's sandbox, check the API still does what its spec says, compare the docs on every release, and set up reconciliation and monitoring so drift gets noticed before a customer notices it.
Delivery and commercial
Most of the hard part sits between the customer conversation and the codebase. I run the technical discovery, take the handoff from sales, build a plan where every task has an owner, run the go-live checks, help sort out incidents and review usage afterwards.
Payments and digital assets
I work on pay-ins, payouts, fiat onramps and wallet flows, where the question is always which status counts as settled and what's allowed to change it. On the chain side, I read on-chain data reliably and handle the sender and recipient details the Travel Rule requires.
Agentic tooling
Agents are good at drafting and bad at knowing what they don't know, so every claim one of my skills makes says whether it came from the docs, a test or an inference, and anything that matters waits for a person. I work with skills, subagents and MCP, which is how an agent connects to your other tools, and I test what a skill produces before anyone relies on it.
Want to talk it through?
Tell me what the system is, where it's stuck and when it needs to be live. A short call usually tells us both whether I'm the right person for it.