AGENTIX · WORK / SYSTEMS INDEX4 SYSTEMS · 10 DOCUMENTED BUILDS / 2026
§ The systems index

Everything we've built,
with its real status.

Not a portfolio of demos. Every entry below states its real status, and is owned outright by the client. Scan the metrics, then open the build.

0custom systems
0documented builds
0%ship with an eval
0named clients
DEPTH OF RECORD
~3,953 commits26+ repos4 custom systems1 in production10 documented builds
NAMED CLIENTS
  • StubsAI
  • A&A
  • Cameron / Elite Capital
§ Track record

Every system is scored
against an eval.

No system reaches production on a hunch. These are the scores, visualised, not asserted.

TAFI279/279
Orions AI55/55
Eval pass-rate on internal story suites
§ Custom systemsengineering programs · multi-week builds
01

TAFI

Natural-Language App Builder
SaaS Platform Development · enterprise software
0/279eval stories · QA 100Certified in internal test runs

A developer types a plain-English app description. TAFI generates the schema, queries, screens, workflows and routing, making 96 live API calls per build in dependency order, with no manual intervention.

Agentic PipelineCrewAIFastMCPQdrant
HOW IT WORKS
Plain EnglishSchema + queries96 API callsRegistered app
Read the full build →
02

Orions AI

Incident Diagnosis Engine
Ticket Marketplace / Operations · StubsAI
0/55epic-01 storiesProduction · read-only

Orions reads a ticket, cross-references StubHub Pro audit logs, MongoDB collections and source code, and publishes an evidence-backed RCA to Azure DevOps. No analyst touches it. Strictly read-only, enforced at the framework level.

Multi-AgentCrewAIQdrant RAGRegression Gate
HOW IT WORKS
Ticket reportCross-ref logs + codeEvidence-backed RCAAzure DevOps
Read the full build →
03

VistaGPT

Construction ERP Intelligence
Construction ERP · in development
0verified schema entries · 4 SQL enginesIn development

Query Viewpoint Vista ERP in plain English. Deterministic ChromaDB stages run before any LLM writes SQL, and a reconcile firewall drops any field not in the database. Four SQL engines sit behind one grounded interface. Internal validation only so far.

AI Data IntelligenceChromaDBLiteLLMPower BI
HOW IT WORKS
Question6 ChromaDB stagesValidated T-SQLPower BI
Read the full build →
04

AI CompMate

The Thinking Layer
Organisational Operations / HR · enterprise client
0 agents9 layers · in active developmentIn active development

AI CompMate passively reads Teams, meetings, email and calendars to build a live causal model of the business. Every morning, each employee gets a brief. No surveys, no manual input, the leadership layer that runs 24/7.

Org AI OSLangGraphNATS JetStreamNeo4j
HOW IT WORKS
Teams · email · calendarLive causal modelDaily brief / employee
Read the full build →
§ Have a system in mind?

Four systems, ~3,953 commits,
one engineering practice.

Each program above started as a process someone wanted automated or data someone wanted to query. Tell us yours, we'll scope the build and the eval it would have to pass.

§ Every engagement includes
Eval harness

A pass/fail suite that must pass before any production deploy.

CI regression gate

Merges blocked the moment quality drops, no silent decay.

100% ownership

Self-hosted. Every line of code is yours from commit one.

Start a project

Your system could be
the next entry here.

Tell us the process you want to automate or the data you want to query. We'll tell you how we'd build it, what it would take, and the eval it would have to pass.

AGENTIX TECH · every system shipped with a score you can audit