Skip to content
Agent-readable

Connect it to your agent.
Ask what I do.

A small knowledge base about my work and the way I build. Point your AI client at it and ask. It answers from cited facts, and it refuses, cleanly, when the answer is not there.

https://alicantorun.com/mcp

What this is

The same architecture I build, pointed at my own work.

Retrieval-first. It returns answers grounded in cited facts from a versioned store, with the source for each, and it declines when a question is out of its corpus. The same pattern I build into products: an agent proposes, a verifier and a human decide. A reference build, not a product.

It answers

  • What I build, and the way I build it
  • How I architect agentic systems, with real scenarios
  • The Lab, the drill, and how I work
  • How I scope, and how to start a conversation

It refuses

  • Anything it cannot ground in my published material
  • It says it does not know, instead of inventing an answer
  • No fabricated numbers, no claims I have not shown
  • The refusal is the point, not a limitation
What your agent can call

A small, growing tool surface.

The tools return cited facts, never a canned answer; the prompts run your agent through a workflow on top. Either way your agent does the reasoning, and the set is on the bleeding edge and grows often, so expect it to change.

ask
tool

Ask anything about my work and how I build. Grounded in cited facts, or an honest refusal.

get_architecturenew
tool

Pull one of my production architecture patterns as structured, actionable guidance.

critique_my_rag_system
prompt

A slash command: your agent critiques your own RAG or agent system against my pattern.

trust_auditnew
tool

My Trust Audit: the principles an AI or software system must pass to be trusted in production, each with the anti-pattern it prevents.

check_fit
tool

Describe your project and get an honest in-scope or out-of-scope read.

search_work
tool

Search my case studies by problem or technology.

get_capabilities
tool

What I do, across Build, Advise, and Empower.

get_drill_scenario
tool

One of 24 real agentic-AI architecture scenarios, with the answer I would give.

get_contact
tool

How to reach me and start a conversation.

On the bleeding edge

New tools land here first. The surface grows often and will change.

Add it

One URL. Add it to your client.

No signup. Adding it is a deliberate choice, and the server only answers questions, it never reaches out on your behalf.

  1. 1Copy the URL or config
  2. 2Add it in your client
  3. 3Ask it about me
terminal
claude mcp add --transport http --scope user alican https://alicantorun.com/mcp

Run it in your terminal. --scope user makes it available in all your projects.

Suggested first prompt

“would alican be a fit for what i’m building, and what would he push back on?”

When it says no

A refusal is a filter, not a dead end.

The server answers only what I have published, and check_fit will call something out of scope when it does not map to my work. That protects you from a confident maybe. But an honest “not covered” is not a no from me. If your problem is close, or you are not sure, reach out and I will look at it myself. Both the refusal and the out-of-scope read hand your agent my booking link and email for exactly this.

Later
MembershipNot open yet

A deeper layer, behind the same server.

Architecture references, starters, and build-faster workflows for agentic startups, behind a membership. Same retrieval, same refusals, more depth.

Prefer a human?

The server is the fast way to find out if a call is worth our time. When it is, let us talk.