S2T Technology · Product direction in development
Tools that do the work and show their work.
S2T Technology is the product company of the S2T family. We are turning patterns from our consulting work into software: assistants that use connected tools, attach evidence to every output, and leave the consequential decisions to a person.
In development. Nothing on this page is for sale yet; early conversations are welcome.
- I · Useful before impressive
Each tool starts with one job a person actually has, and is judged by whether that job got easier.
- II · Evidence in the output
Every result carries where it came from, what checked it, and what is uncertain.
- III · A person at the boundary
Tools prepare and propose. Anything with consequences waits for a human to approve.
- IV · Connected on purpose
Tools reach other systems through explicit, scoped connections such as the Model Context Protocol, not through hidden access.
The product thesis
Automation moves the work. Good software keeps the reason it worked.
A one-off fix solves today's case. A product carries the pattern forward: the context it needs, the limits it respects, where its answers came from, and how to check them. That is what turns one solution into something another person can use.
What we're building
Four product areas, one set of rules.
Each area starts from a different need. All follow the same rules: bounded jobs, visible evidence, a person at the boundary, and connections made on purpose.
- I
Tool connections
A tool server using the Model Context Protocol that gives AI assistants a scoped set of tools: text embeddings for search, CloudFormation template generation, and OAuth configuration checks. The code is public at github.com/S2TConsulting/accelerators-mcp.
- II
Desktop assessment client
A desktop application for running technology assessments against a cloud estate and turning the findings into organized, reusable records. In private development.
- III
Pattern library
Implementation patterns from real delivery work, packaged with scripts and procedures so the next engagement starts ahead. Used internally today; packaging for others is in progress.
- IV
Dashboards for agent work
Views that show what automated agents are doing, what they are waiting on, and the evidence they produced: a relay dashboard and an observatory for multi-agent work. In private development.
Status
Direction in development.
Products described on this page are in development and are not offered for sale.
How we build
Start useful. Earn complexity.
Each layer is added only after the one before it has proved its value in use.
- 01
Distill
Separate the durable pattern from the one-time circumstances that revealed it.
- 02
Encode the evidence
Make provenance, checks, confidence, and limits part of how the tool behaves, not a note in the manual.
- 03
Design for the job
Build for the task a person is actually trying to finish, measured in time and errors saved, not in how the system looks.
- 04
Compose with care
Connect tools only where the connection serves a job, and keep each one understandable on its own.
Built with today's AI
Agentic, connected, accountable.
Current AI models can plan multi-step work, call tools, and act on real systems. We build on those capabilities deliberately: agents run inside written gates, reach systems through scoped tool connections such as the Model Context Protocol, and produce outputs that carry their own evidence. The same rules govern how we build: AI assistants write and review code here under human approval.
- Agentic patterns
Multi-step work is broken into bounded tasks with a visible state, not a single opaque run.
- Model Context Protocol
The tool server exposes capabilities to assistants through a standard interface, scoped to what each task needs.
- Evidence attached
Each result records the inputs, the command or model call, the checks, and what remains uncertain.
- Human approval
Sends and deployments wait for a person. The gate is enforced in code, not only in a policy document.
The companies · distinct mandates
Products with a place in the system.
Technology is the product company of a small family. Each organization has one job.
-
I
Thomas Capital Holdings
Holding company. Holds the shared intellectual property and keeps the companies separate.
-
II
S2T Consulting
Operating consulting firm. Where the patterns we turn into tools come from.
-
III
S2T Technology
Product company. This site. Direction in development.
-
IV
S2T Foundation
Independent public-benefit initiative in formation. Practical access and learning. Not owned by the commercial companies.
Bring us the problem you keep solving by hand.
If the same job keeps getting done with a workaround, tell us about it. We are choosing what to build next from problems like that, and early conversations shape the tools.
shaun@s2ttechnology.comResearch / Full-length editions
Open the architecture. Inspect the argument.
Technical papers on agentic infrastructure, reducing system complexity, and preparing governed work before execution.
Agentic Infrastructure
A reference design for connecting identity, evidence, reusable capabilities, and accountable automation.
The Subtraction Principle
Reducing unnecessary complexity while preserving useful capability, recovery, security, and evidence.
The Workflow Compiler
Preparing bounded, versioned work packets before execution while keeping live authorization checks at the point of action.
The full texts preserve the original reasoning, technical detail, and references. Company and client identities use disclosed aliases; confidential details are withheld. Later revisions and editorial notes carry separate dates.