> ## Documentation Index
> Fetch the complete documentation index at: https://developers.lighton.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Tools

> Reference for all tools exposed by the LightOn MCP server.

The LightOn MCP server exposes three tools. Use `list_workspaces` and `list_tags` to resolve names to IDs, then pass those IDs to `search` to scope your queries.

***

## search

Retrieve the most relevant passages from your indexed documents for a natural-language query.

<ParamField body="query" type="string" required>
  Natural-language search query. Must be self-contained: resolve pronouns and inline any necessary context. Max 1500 characters.
</ParamField>

<ParamField body="workspace_ids" type="integer[]">
  Restrict the search to these workspace IDs. Cannot be combined with `file_ids`.
</ParamField>

<ParamField body="file_ids" type="integer[]">
  Restrict the search to specific file IDs. Cannot be combined with `workspace_ids` or `tag_ids`.
</ParamField>

<ParamField body="tag_ids" type="integer[]">
  Restrict the search to documents carrying any of these tag IDs. Cannot be combined with `file_ids`.
</ParamField>

<ParamField body="max_results" type="integer" default="10">
  Maximum number of chunks to return. Range: 1–50.
</ParamField>

<ParamField body="mode" type="string" default="text">
  Search mode. `"text"` uses hybrid keyword + vector search. `"vision"` uses VLM-embedded page image search.
</ParamField>

***

## list\_workspaces

List every workspace the caller has access to, each annotated with an auto-generated summary. Use this to resolve a workspace name to an ID before scoping a `search` call.

No parameters.

***

## list\_tags

List every document tag defined within the caller's company. Use this to map a human-readable label to a `tag_id` for filtering searches.

No parameters.