Do we need a trained MetatextAI model before Lyro can classify anything?

For classification, yes. Classify Text returns labels and confidence scores from a model trained in your MetatextAI project, so the label set is yours rather than a generic taxonomy. List Models shows which models exist and what tasks each supports, and Lyro reads that before choosing one.

Which application do the guardrails Lyro checks against apply to?

Policies are scoped per application, not per account. Lyro calls List Applications to resolve the application ID first, then List Guardrail Policies for that ID, so a rule created on one application is not silently applied to another.

Can Lyro change or remove our guardrail policies?

The toolkit exposes Create Policy Guardrails, Update Policy Guardrails, and Delete Guardrail Policy, so technically it can. Those calls change what every downstream check enforces, so treat them as an admin operation and scope the connected key accordingly instead of leaving them reachable from a customer-facing conversation.

How is this different from putting a content rule in the prompt?

A prompt instruction is advisory and quietly disappears when the wording is edited. Evaluate Messages checks finished output against the policies stored in MetatextAI and returns violation details or a corrected version, so Lyro acts on a verdict rather than assuming the model complied.

Metatextai MCP integration

Labels incoming text, extracts structured fields, and screens drafted replies against the guardrail policies your application already enforces.

Connect Metatextai See all integrations

13actions available

Three actions you can hand over today

Every action runs live through MCP. Nothing to build, nothing to maintain.

Chat completions
Lyro generates responses that are automatically guardrailed for safety. Customers get accurate answers while your team avoids harmful or policy-violating outputs.
Evaluate messages
Lyro validates responses against safety policies before sending. Your team catches risky content before it reaches customers, protecting brand reputation.
Update policy guardrails
Lyro adjusts safety rules for different customer scenarios. Your team enforces stricter guardrails for sensitive issues or loosens them for casual conversations.

See all 13 actions

How businesses use Metatextai + Lyro

Each card is one request a support team gets, and the Metatextai actions Lyro runs to close it.

Turn a free-text message into structured fields
Lyro classifies the message against a trained MetatextAI model and runs extraction over the same text, so an intent, a date, or an order number arrives as labelled data instead of a paragraph to read.
Classify TextRun ExtractionList Models
Draft a reply and screen it in one pass
Lyro generates the wording with a project model, then evaluates the result against your policies and gets back violation details or a correction before the message is ever shown to a customer.
Generate TextChat CompletionsEvaluate Messages
Keep guardrail policies current
Lyro lists the policies attached to an application, adds a new rule, or amends an existing one, so the constraints your content has to satisfy live in one place rather than scattered across prompts.
List Guardrail PoliciesCreate Policy GuardrailsUpdate Policy Guardrails
Red-team an application against known probes
Lyro lists the available red team probes and runs a vulnerability scan against a named application on request, so a security check is something a person asks for rather than a job someone has to schedule.
List Red Team Test ProbesRun Red Team Test ScanList Applications

How it works

Get started in 3 steps

Connect once, then just ask. There is no workflow builder to learn and nothing to maintain — Lyro reads the Metatextai actions it has and picks the ones a request needs.

01
Connect Metatextai
Authorize the Metatextai account your team already uses — one consent screen, no API keys, no mapping tables. Lyro can only do what you granted that account, and you can disconnect it at any time.
02
Tell your agent what you need
Describe the job the way you would hand it to a teammate. Lyro maps it to the Metatextai actions that close it and chains as many as the request needs.
03
Watch it work
The agent runs the actions inside the conversation the customer is already in, so nobody copies data between tabs and your team can take over at any point.
Get started free

Metatextai · Lyro

Everything else about Metatextai

Setup, permissions, and the limits of what Lyro can do inside Metatextai.

For classification, yes. Classify Text returns labels and confidence scores from a model trained in your MetatextAI project, so the label set is yours rather than a generic taxonomy. List Models shows which models exist and what tasks each supports, and Lyro reads that before choosing one.