Claude Code Skills

notery-search

Search your Notery workspace documents using semantic search via MCP.

Overview

The notery_search tool performs semantic search across all documents in your workspace. It uses Cohere embeddings and pgvector to find relevant documents even when the exact keywords do not appear in the text.

This tool is available through both the MCP server and agent skills. It calls the underlying REST endpoint:

GET/api/v1/search

Semantic search across workspace documents. See API Reference for direct HTTP usage.

Parameters

Parameter	Type	Required	Description
`query`	string	Yes	The search query. Can be a keyword, phrase, or natural language question.
`limit`	number	No	Maximum number of results to return. Default: `10`. Max: `50`.

Response

The tool returns formatted text containing a list of matching documents, ranked by relevance. Each result includes:

Title — the document title
Path — the full path within the workspace
ID — the document UUID (use this with notery_get to fetch full content)
Similarity score — how closely the document matches the query
Tags — any tags assigned to the document
Summary / Excerpt — the document summary if available, otherwise a content excerpt

Pro plan users get reranked results. After the initial vector search, a reranking model reorders the results for better relevance, especially for nuanced or multi-part queries.

Example

Ask your AI assistant:

"Search my Notery docs for authentication flow"

The tool will return something like:

Found 3 results for "authentication flow":

1. Auth Architecture
   Path: /backend/auth-architecture
   ID: 550e8400-e29b-41d4-a716-446655440000
   Score: 0.92
   Tags: auth, backend, security
   Summary: Overview of the BetterAuth setup including session management,
   email/password flow, and post-signup workspace provisioning.

2. Login Page Implementation
   Path: /frontend/login-page
   ID: 6ba7b810-9dad-11d1-80b4-00c04fd430c8
   Score: 0.87
   Tags: frontend, auth
   Summary: Implementation notes for the login form component using
   Nuxt UI's UAuthForm with email/password fields.

3. API Key Authentication
   Path: /api/api-key-auth
   ID: 7c9e6679-7425-40de-944b-e07fc1f90ae7
   Score: 0.71
   Tags: api, auth
   Summary: How API keys are validated in the /api/v1/* middleware
   and scoped to workspaces.

Typical Workflow

Use notery_search to find documents matching your question.
Review the titles, paths, and summaries to identify the most relevant result.
Use notery_get with the document ID to retrieve the full content.

This two-step pattern keeps initial responses fast by returning summaries first, then fetching full content only when needed.