Products & Connections · Beginner tutorial01
How to set up Products & Connections in Truvyx
Create a product profile, separate development, staging, and production, choose a connection method, and verify the first safe request.
Read guideBlog / How-to guides
Learn Truvyx by doing.
You should not need to understand evaluation theory before you can use the platform. These guides walk through the practical path: what to click, what to write, what the result means, and what to do next.
Ground truth · Methodology02
How to create ground truth for complex AI workflows
Turn decisions, constraints, handoffs, and acceptable variations into evidence an AI evaluation can check.
Read guideAI metrics · Framework03
Feasibility, completeness and optimality: a better AI scorecard
Learn why three complementary dimensions reveal more than one accuracy number when you evaluate an AI agent.
Read guideVerifier scripts · Technical tutorial04
How to build verifier scripts for LLM and agent outputs
Write small, clear, repeatable checks for fields, constraints, tool calls, traces, and evidence.
Read guideRoot cause analysis · Pillar guide05
Root cause analysis for AI agents: architecture and workflow
Move from a failed run to an evidence-backed cause across prompts, tools, handoffs, constraints, and traces.
Read guideDiagnosis · Technical guide06
Counterfactual diagnosis: finding the smallest fix for an AI failure
Change one variable at a time, compare the evidence, and promote the smallest verified repair into a regression test.
Read guideMCP · Beginner tutorial07
How to use Truvyx from Claude Code, Cursor, or another MCP client
Create a scoped key, connect the Truvyx MCP server, inspect its manifest, investigate runs, and trigger safe test workflows from your editor.
Read guideSDK Docs · Beginner tutorial08
How to add Truvyx evaluation to an AI agent
Install the TypeScript or Python SDK, run an evaluation, read EvalResult, add CI gates, capture traces, and protect multi-agent contracts.
Read guideCost Benchmarks · Beginner tutorial09
How to measure AI cost and efficiency in Truvyx
Understand model pricing, token usage, redundant calls, cost per correct decision, and efficiency benchmarks.
Read guideHITL Analytics · Beginner tutorial010
How to use Truvyx HITL Analytics to improve human-agent workflows
Measure human-agent friction, find correction patterns, and improve escalation workflows with evidence.
Read guideChaos Studio · Beginner tutorial011
How to test AI resiliency with Truvyx Chaos Studio
Create controlled MCP fault-injection experiments, score agent recovery, and export resiliency evidence.
Read guideInvestigate · Beginner tutorial012
How to investigate AI behavior with Truvyx Investigate
Ask grounded questions about runs and RCA findings, then track recurring failures automatically.
Read guideRCA Engine · Beginner tutorial013
How to investigate an AI failure with Truvyx RCA Engine
Trace a failed run to its root cause, test counterfactual fixes, and create a verified remediation plan.
Read guideRegistry · Beginner tutorial014
How to use the Truvyx Registry: a beginner's guide
Find, inspect, download, cite, reuse, and contribute validated AI evaluation scenarios.
Read guideDecomposition Designer015
How to use Truvyx Decomposition Designer to map multi-agent systems
Design agent nodes, dependencies, information boundaries, contracts, anti-pattern checks, and framework exports from one visual canvas.
Read guideVerifier Factory016
How to use Truvyx Verifier Factory to test AI outputs
Generate, inspect, test, download, and run versioned verifier scripts, then connect evidence to Evaluation Runs and CI/CD.
Read guideConstraint Engine017
How to use Truvyx Constraint Engine for safer AI evaluations
Define hard and soft constraints, check feasibility, configure escalation, protect sensitive fields, and connect rules to verification and diagnosis.
Read guideScenario Studio018
How to use Truvyx Scenario Studio for your first AI evaluation
A beginner-first walkthrough of the real Scenario Studio workflow: domains, templates, Natural Language generation, constraints, saving, and the next module to open.
Read guideStart here019
How to run your first AI evaluation in Truvyx
Create a scenario, submit an agent output, and understand what PASSED, FAILED, and PARTIAL mean.
Read guideScenario design020
How to write a useful evaluation scenario
Turn a vague expectation into constraints, ground truth, and checks that can produce a useful diagnosis.
Read guideAPI setup021
How to connect your agent with the Truvyx API
Use an API key, submit camelCase evaluation fields, poll the result, and handle failures safely.
Read guideResults022
How to read an evaluation result
Learn the difference between a score, a violation, a completeness gap, and a root-cause signal.
Read guideScaling up023
How to move from one test to continuous evaluation
Build a small regression suite and add evaluation to the development workflow without creating noise.
Read guideThe simplest Truvyx mental model
1. Scenario: describe what good behaviour looks like.
2. Run: submit what your agent actually did.
3. Diagnose: fix the smallest cause the evidence points to.