Glossary

structuredContent

structuredContent is the machine-readable part of an MCP tool result, shaped by a schema so clients can read fields instead of parsing prose.

Updated November 12, 2026

Standard
Model Context Protocol specification
Also called
structured tool output

At OneLifeVibe

What it means

structuredContent is the part of a tool result that code can read directly. The rest is usually human-readable content: text that tells the person what happened. structuredContent carries the same outcome as fields, and a schema says which fields exist and what types they use.

That split matters when the result changes what the client does next. Reading a field like error is less brittle than matching a phrase in a message. The schema gives both sides a contract: the server knows what to return, the client knows what to expect. If a client cannot use the schema, it can still show the text.

Keep the two parts consistent. The text can be short because the fields carry the details.

How it works

An MCP server declares a schema for a tool's output, then returns structuredContent that follows it. The client reads the result, checks isError, and validates the fields before using them. Types apply: strings stay strings, booleans stay booleans.

A tools/call result can include both content and structuredContent. content is for display. structuredContent is for branching, logging, or passing values to another step. If the schema is missing or validation fails, treat the text as the fallback rather than guessing at field names.

Some failures are transport failures: the request did not complete. Others are outcomes the person can fix, such as a missing link. Those can come back as a normal result with isError set to true and structuredContent that names the problem. Do not retry those blindly.

How OneLifeVibe uses it

OneLifeVibe uses that shape for tool failures a person can fix. For example, if a website is not linked yet, the tool answers a normal result with isError: true and structuredContent like {"error":"not_linked","site":"helpbless.com"}. The error field carries a code such as not_linked. The site field names the website involved.

The client should branch on error, not on the wording of the message. A not_linked code means the person needs to link that site before the same call can work. The site value tells the client which site to use in that step.

Marking the result isError: true lets the client show a failure state without treating the call as a broken connection. Do not throw away structuredContent just because isError is true. That is when the fields are most useful.

Common mistakes

The first mistake is parsing the human-readable message to find out what failed. The message may change; error is the field to read.

The second is treating every isError: true result as a transport failure. A normal result with isError: true is still a completed call. Inspect the code, then decide whether the person can fix it.

The third is assuming structuredContent is on every tool result. Tools with no schema for structured output may only return content. Check for the field and keep a text fallback. Validate types too: a schema gives the shape, but your client still has to handle values it does not recognize.

Questions

What is the difference between content and structuredContent in MCP?

content is the human-readable part of a tool result. structuredContent is the machine-readable part, shaped by a schema. Clients can show content and use structuredContent for decisions.

Why does a OneLifeVibe tool return isError true with structuredContent?

It does that when a tool fails for a reason the person can fix, such as a website that is not linked yet. The structuredContent names the error and the site, for example {"error":"not_linked","site":"helpbless.com"}. The call completed normally, so retrying the same request will not help until the site is linked.

See it working. The OneLifeVibe connector and Platform API put these pieces together.

Get a free key Read the docs

Written from the OneLifeVibe tool registry and platform configuration, and checked automatically before publishing. Something wrong? Tell us.