get_item

Tool

get_item

Capability: read Rate limit: 1000 / hr per token

Parameters

Example call

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_item",
    "arguments": { "item_id": "74b18cbe-20b1-494d-96bf-64c5715f4a6f" }
  }
}

Example response

The tool returns formatted markdown that includes the item's title, platform, source URL, AI transforms (if generated), and the full transcript or source text.

# Andrej Karpathy on the future of training
Platform: youtube
Source: https://www.youtube.com/watch?v=...

## Auto transform
_The agent's take on a 1h42m talk in 320 words..._

(transform body here)

## Transcript / source text
(full transcript)

The structuredContent mirrors this with typed fields: outputs, url, title, platform, and others.

Scope behavior

The item must belong to the caller. For scoped tokens, the item must also be a member of at least one folder in the token's scope. Out-of-scope requests return a not_found error rather than forbidden_scope, so the existence of out-of-scope items is not leaked.

kb_only tokens additionally reject items flagged as not-in-KB.

Errors

Related

• search_knowledge_base: find items by query first, then fetch with get_item
• REST /api/v1/items/{id}: the same item data with a stable JSON shape (preferred for non-agent clients)