MCP tools: Assistance

Last updated About 2 hours ago

Tools that neither read nor write project data, but help the AI answer: product knowledge, the support path, and how things are rendered in chat.

`search_docs`

Searches the Liza documentation (what Liza is, how it works, the Public API and MCP). read · guest

Parameter	Type	Default	Description
`query`	string	–	Search term. Without `query` you get the documentation overview (index)
`limit`	number	`5`	Max number of matching sections

The source is the very pages you are reading — the server ingests them as llms.txt/llms-full.txt from docs.liza.app.

Creates a post in Featurebase (feedback.liza.app) — the path for bug reports, feature requests, feedback and support.

write · user

Parameter	Type	Description
`type`	enum `bug` \| `feature` \| `feedback` \| `support` (required)	Kind of post
`title`	string (3–120) (required)	Short, clear title
`content`	string (5–6000) (required)	Description in markdown — for bugs include steps to reproduce, expected vs. actual behaviour

The response contains the real post URL (feedback.liza.app/p/…). Do not create an internal list item and never invent a ticket ID.

Returns the specification of the Liza chat widgets. read · guest

Widgets are fenced code blocks that Liza renders in chat as real elements instead of text:

Block	Result
```liza-chart	Chart (bar, line, pie …)
```liza-flow	Flow diagram
```liza-form	Fillable form
```liza-action	Action button that really writes after confirmation

Parameter	Type	Default	Description
`type`	enum `overview` \| `chart` \| `flow` \| `form` \| `action`	`overview`	`overview` = which widget when + shared rules; otherwise the full schema of that type

Call it before writing the first such block in a session instead of guessing the JSON schema.