Core Tools Reference

Every Trove connection ships with 8 Core tools. All tool calls use the tools/call JSON-RPC method with the tool name and arguments.

Source names vs IDs: the Core tools accept source names (e.g., "Readwise", case-insensitive) for the source parameter. The GraphQL API uses sourceId (the source’s ID) instead. Use trove_list_sources to see available source names.

trove_search

Semantic search across your knowledge base. Covers bookmarks, highlights, articles, transcripts, and saved content from all your sources.

Parameters

Parameter	Type	Required	Default	Description
query	string	Yes		Natural language search query
source	string	No		Filter by source name (e.g., "Readwise")
feedId	string	No		Filter to a specific feed by ID
author	string	No		Filter by content author
after	string	No		ISO 8601 date. Only content after this date
before	string	No		ISO 8601 date. Only content before this date
limit	number	No	10	Max results (max 25)

Example request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "trove_search",
    "arguments": {
      "query": "transformer architecture attention mechanism",
      "limit": 5
    }
  }
}

Example response

The result.content[0].text field contains formatted text:

Found 5 results (45ms):

1. [0.92] Attention Is All You Need
   Source: arXiv Papers | Author: Vaswani et al. | 8,432 words
   "The dominant sequence transduction models are based on complex recurrent..."
   [doc:a1b2c3d4]

2. [0.87] The Illustrated Transformer
   Source: Readwise | Author: Jay Alammar | 3,201 words
   "In this post, we will look at The Transformer..."
   [doc:e5f6g7h8]

Tips

• Be specific. "attention mechanism in transformers" outperforms "AI stuff".
• Use the source filter to scope searches to a single source.
• Use date filters (after, before) to narrow results by time.
• Use trove_get_document to read the full text of any result.

---

trove_get_document

Retrieve the full text of a specific document by ID. Use this after trove_search to read complete content.

Parameters

Parameter	Type	Required	Description
document_id	string	Yes	Document ID from search results (the [doc:...] reference)

Example request

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "trove_get_document",
    "arguments": {
      "document_id": "a1b2c3d4"
    }
  }
}

Example response

The response contains formatted text with the document title, source, author, date, word count, and the full content body.

Tips

• Only fetch full text when the snippet from search results is not enough. Full text is loaded from object storage and is larger.
• The document_id is the value inside [doc:...] brackets in search results.

---

trove_get_source

Detailed information about a source, including document count, top authors, recent documents, and sync history.

Parameters

Parameter	Type	Required	Description
source	string	Yes	Source name (e.g., "Readwise") or ID

Example request

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "trove_get_source",
    "arguments": {
      "source": "Readwise"
    }
  }
}

Response includes

• Source type, status, and collection shape (kind/transport)
• Total document count
• Last sync time
• Date range of indexed content
• Feeds within the source, each with its sync position (watermark)
• Top 5 authors by document count
• 5 most recent documents
• Last 5 sync runs with status

Tips

• Use the source name (case-insensitive) rather than the ID for convenience.
• Call trove_list_sources first if you are not sure of the source name.

---

trove_recent

Recently indexed content in chronological order (newest first). A timeline view, not a semantic search.

Parameters

Parameter	Type	Required	Default	Description
source	string	No		Filter by source name
feedId	string	No		Filter to a specific feed by ID
author	string	No		Filter by author
since	string	No	last 24h	ISO 8601 datetime. Only content indexed after this time
limit	number	No	15	Max results (max 50)

Example request

{
  "jsonrpc": "2.0",
  "id": 4,
  "method": "tools/call",
  "params": {
    "name": "trove_recent",
    "arguments": {
      "since": "2026-03-24T00:00:00Z",
      "limit": 10
    }
  }
}

Tips

• No parameters required. Calling with no arguments returns the last 15 items indexed in the past 24 hours.
• Good for “what’s new?” queries and reviewing recently saved content.

---

trove_list_documents

Exhaustively list documents with filters. Unlike trove_search (top-K semantic matches), this returns an authoritative totalCount, so “all X” questions get a complete, verifiable answer. Page through large lists with offset.

Parameters

Parameter	Type	Required	Default	Description
source	string	No		Filter by source name (e.g., "Apple Podcasts")
author	string	No		Filter by author or show (fuzzy, case-insensitive)
content_type	string	No		Filter by type: text, transcript, bookmark, or highlight
after	string	No		ISO 8601 date. Only documents published after this date
before	string	No		ISO 8601 date. Only documents published before this date
sort	string	No	newest	Sort order: newest, oldest, recently_indexed, or title
limit	number	No	25	Page size (max 50)
offset	number	No	0	Skip this many documents (for paging)

Example request

{
  "jsonrpc": "2.0",
  "id": 5,
  "method": "tools/call",
  "params": {
    "name": "trove_list_documents",
    "arguments": {
      "source": "Apple Podcasts",
      "sort": "newest",
      "limit": 25
    }
  }
}

Tips

• Use this (not trove_search) for counts and “list every X” requests — it reports an authoritative totalCount.
• For podcasts, the show name is stored as the document author, so filter by author to scope to a single show.
• Page through results with offset when the total exceeds the page limit.

---

trove_discover

Broad exploration of the knowledge base around a topic. Unlike trove_search, which returns precise matches, trove_discover uses a lower similarity threshold to surface loosely related content across sources.

Parameters

Parameter	Type	Required	Default	Description
topic	string	Yes		Topic or theme to explore
source	string	No		Filter by source name
feedId	string	No		Filter to a specific feed by ID
limit	number	No	10	Max results (max 25)

Example request

{
  "jsonrpc": "2.0",
  "id": 5,
  "method": "tools/call",
  "params": {
    "name": "trove_discover",
    "arguments": {
      "topic": "decision-making under uncertainty"
    }
  }
}

Tips

• Use for brainstorming sessions and exploring adjacent ideas.
• More exploratory than trove_search. Useful when you want surprising connections across sources.
• Pair with trove_get_document to dive deeper into anything interesting.

---

trove_list_sources

List all your sources and their current status.

Parameters

None.

Example request

{
  "jsonrpc": "2.0",
  "id": 6,
  "method": "tools/call",
  "params": {
    "name": "trove_list_sources",
    "arguments": {}
  }
}

Example response

The response lists each source with its status:

3 sources:

- Readwise (readwise, api) [active]
  Documents: 1,247
  Last synced: 2026-03-25T08:14:00Z

- Kindle Highlights (kindle, local) [active]
  Documents: 342
  Last synced: 2026-03-24T06:30:00Z

- Manual Saves (manual) [active]
  Documents: 28

The parenthetical shows the source type and its transport (and kind when on-demand). Status values: [active], [paused], [error], [setup]. Sources with [error] status include an error message.

Tips

• Call this first to see what sources are available before searching.
• Source names from this list can be used as the source filter in other tools.

---

trove_save

Save content from the current conversation into your knowledge base. Accepts text directly or a URL to fetch and index.

Parameters

Parameter	Type	Required	Description
text	string	No*	Text content to save
url	string	No*	URL to fetch and index
title	string	No	Title for the saved content
source	string	No	Attribution (e.g., "conversation with Claude")
tags	string[]	No	Categorization tags

*At least one of text or url must be provided.

Example request

{
  "jsonrpc": "2.0",
  "id": 7,
  "method": "tools/call",
  "params": {
    "name": "trove_save",
    "arguments": {
      "text": "The Pareto principle states that roughly 80% of consequences come from 20% of causes.",
      "title": "Pareto Principle",
      "source": "conversation with Claude",
      "tags": ["mental-models", "decision-making"]
    }
  }
}

Example response

The response confirms the save:

Saved: "Pareto Principle"

Document ID: f9a8b7c6
Words: 16
Source: conversation with Claude
Tags: mental-models, decision-making

The content is now searchable via trove_search.

Tips

• Saved content goes to the Manual Saves source, which is auto-created on first use.
• Content is immediately searchable via trove_search after saving.
• Use tags to organize saved content for easier filtering later.
• The source field is free-form attribution — it helps you remember where the content came from.
• When saving a URL, Trove fetches the page content. You do not need to paste the text yourself.

See Also

The Core tools map to Trove’s GraphQL queries and mutations. The GraphQL API offers additional filtering and pagination options not available through the connection.