Lightdrift Images MCP

Search images

search_images

Search the Voyage index of 1,852,370 captioned images using joint image-and-caption embeddings, hybrid retrieval and caption text reranking. Supports text, uploaded images via image_base64, public image_url, or image plus text. Supply query text or one image; icon-pack search is not supported. Describe subjects, recognizable names, visible text, mood, style or composition in natural language, such as 'misty mountains at sunrise'. Use find_similar_images when you already have a Lightdrift asset_id, or get_image to retrieve one known image without a search. Returns query_id and a ranked results array with asset_id, file and thumbnail URLs, source, dimensions, and rights including attribution. Set k for result count; optional candidate and ranking controls normally need no changes. Defaults exclude AI-generated images, require commercial-use licenses, and require width >= 1000 pixels; nsfw_max defaults to 0.2. These four defaulted filters accept null to opt out; unknown NSFW scores pass. Requires OAuth or an API key and is subject to account rate limits and available credits. Each successful call charges $0.02 and records usage; repeated calls can charge again. Does not modify images or publish anything. Inspect each result's rights before use.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`k`	No	Maximum number of final results (1–100, default 10). Fewer results may match the filters; this is separate from candidate-pool sizes.
`mode`	No	Ranking mode: auto uses caption text reranking for text queries and blends image similarity with text reranking for image-plus-text. Image-only uses image similarity. text forces text reranking and requires text. none skips reranking. All modes use the Voyage index.	auto
`ann_k`	No	Maximum semantic-search candidates (1–1000, default 100). Increase for a wider pool; this does not set the final result count.
`query`	No	Optional natural-language query. Combine with an image to describe desired changes, or omit for image-only search.
`ann_ef`	No	Legacy retrieval tuning parameter (64–4096, default 2048). Accepted for compatibility; leave at the default for the current retrieval service.
`bm25_k`	No	Maximum keyword-search candidates (0–1000, default 50), combined with ann_k semantic candidates. Set 0 for semantic-only retrieval.
`explain`	No	Include retrieval and ranking diagnostics with each result (default false). Use for debugging search quality.
`filters`	No	Combine license, source, size, and visual filters. Omitted commercial, ai_generated, min_width and nsfw_max default to true, false, 1000 and 0.2. Set any of these four to null to remove its default.
`rerank_k`	No	Maximum candidates to rerank (1–500, default 60); null reranks the whole candidate pool and can exceed provider limits; prefer a bounded value. Ignored when mode is none. Independent of final result count k.
`image_url`	No	Public HTTPS URL of a JPEG, PNG or WebP image, maximum 5 MiB and 20 megapixels. Private addresses and authenticated URLs are unsupported. Optional query text refines the image search.
`rerank_text`	No	Legacy reranker text toggle (default true). The current service always uses text; leave true.
`image_base64`	No	JPEG, PNG or WebP upload: base64 data URL or raw base64, maximum 5 MiB decoded and 20 megapixels. Use this or image_url, not both.

Schema Changelog

Changes observed during successful MCP inspections.

Changed6 schema fields changedSep 22, 2026

addedInput schema / properties / image_base64

Added value: +{
+  "description": "JPEG, PNG or WebP upload: base64 data URL or raw base64, maximum 5 MiB decoded and 20 megapixels. Use this or image_url, not both.",
+  "maxLength": 6990572,
+  "type": "string"
+}

addedInput schema / properties / image_url

Added value: +{
+  "description": "Public HTTPS URL of a JPEG, PNG or WebP image, maximum 5 MiB and 20 megapixels. Private addresses and authenticated URLs are unsupported. Optional query text refines the image search.",
+  "format": "uri",
+  "maxLength": 4096,
+  "type": "string"
+}

changedInput schema / properties / mode / description

Previous value: -"Ranking mode: auto (default) and text use Voyage caption text reranking; none skips reranking. Every mode uses the same joint image-and-caption index, never a legacy visual reranker."New value: +"Ranking mode: auto uses caption text reranking for text queries and blends image similarity with text reranking for image-plus-text. Image-only uses image similarity. text forces text reranking and requires text. none skips reranking. All modes use the Voyage index."

changedInput schema / properties / query / description

Previous value: -"Describe the image, subject, scene, or mood to find (1–1000 characters), for example: misty mountains at sunrise."New value: +"Optional natural-language query. Combine with an image to describe desired changes, or omit for image-only search."

removedInput schema / properties / query / minLength
```
Removed value: -1
```
removedInput schema / required
```
Removed value: -[
-  "query"
-]
```

Changed10 schema fields changedSep 22, 2026

changedInput schema / properties / filters / description

Previous value: -"Combine license, source, size, and visual filters. Omitted commercial, ai_generated, and min_width default to true, false, and 1000; set those fields to null to remove their defaults."New value: +"Combine license, source, size, and visual filters. Omitted commercial, ai_generated, min_width and nsfw_max default to true, false, 1000 and 0.2. Set any of these four to null to remove its default."

addedInput schema / properties / filters / properties / nsfw_max / anyOf

Added value: +[
+  {
+    "maximum": 1,
+    "minimum": 0,
+    "type": "number"
+  },
+  {
+    "type": "null"
+  }
+]

changedInput schema / properties / filters / properties / nsfw_max / description

Previous value: -"Maximum recorded NSFW score from 0 to 1; lower is stricter. Defaults to 0.2. Assets without a score can still match."New value: +"Maximum recorded NSFW score from 0 to 1; lower is stricter. Defaults to 0.2; null removes the restriction. Assets without a score can still match."

removedInput schema / properties / filters / properties / nsfw_max / maximum
```
Removed value: -1
```
removedInput schema / properties / filters / properties / nsfw_max / minimum
```
Removed value: -0
```
removedInput schema / properties / filters / properties / nsfw_max / type
```
Removed value: -"number"
```
changedInput schema / properties / mode / default
```
Previous value: -"visual"New value: +"auto"
```

changedInput schema / properties / mode / description

Previous value: -"Ranking mode: visual selects the default relevance reranker; none skips reranking and returns retrieval order. Default visual."New value: +"Ranking mode: auto (default) and text use Voyage caption text reranking; none skips reranking. Every mode uses the same joint image-and-caption index, never a legacy visual reranker."

changedInput schema / properties / mode / enum

Previous value: -[
-  "visual",
-  "none"
-]New value: +[
+  "auto",
+  "text",
+  "none"
+]

changedInput schema / properties / rerank_k / description

Previous value: -"Maximum candidates to rerank (1–500, default 60); null reranks the whole candidate pool. Ignored when mode is none. Independent of final result count k."New value: +"Maximum candidates to rerank (1–500, default 60); null reranks the whole candidate pool and can exceed provider limits; prefer a bounded value. Ignored when mode is none. Independent of final result count k."

Changed25 schema fields changedSep 22, 2026

addedInput schema / properties / ann_ef / description

Added value: +"Legacy retrieval tuning parameter (64–4096, default 2048). Accepted for compatibility; leave at the default for the current retrieval service."

addedInput schema / properties / ann_k / description

Added value: +"Maximum semantic-search candidates (1–1000, default 100). Increase for a wider pool; this does not set the final result count."

addedInput schema / properties / bm25_k / description

Added value: +"Maximum keyword-search candidates (0–1000, default 50), combined with ann_k semantic candidates. Set 0 for semantic-only retrieval."

addedInput schema / properties / explain / description

Added value: +"Include retrieval and ranking diagnostics with each result (default false). Use for debugging search quality."

addedInput schema / properties / filters / description

Added value: +"Combine license, source, size, and visual filters. Omitted commercial, ai_generated, and min_width default to true, false, and 1000; set those fields to null to remove their defaults."

addedInput schema / properties / filters / properties / ai_generated / description

Added value: +"Filter by AI-generated status. Defaults to false when omitted; null allows either. False does not exclude non-AI illustrations."

addedInput schema / properties / filters / properties / attribution_required / description

Added value: +"Filter by whether the license requires attribution. Omit to allow either."

addedInput schema / properties / filters / properties / colors / description

Added value: +"Match any listed dominant color name, for example [\"blue\", \"teal\"]. Images without matching color metadata are excluded."

addedInput schema / properties / filters / properties / commercial / description

Added value: +"Require a commercial-use license when true. Defaults to true when omitted; null removes this filter."

addedInput schema / properties / filters / properties / derivatives / description

Added value: +"Filter by whether the license permits modified or derivative works. Omit to allow either."

addedInput schema / properties / filters / properties / format / description

Added value: +"Allowed file formats, for example [\"jpeg\", \"png\"]. Omit to allow any format."

addedInput schema / properties / filters / properties / license_id / description

Added value: +"Allowed license IDs, for example [\"cc0\", \"cc-by\"]. Omit to allow all licenses compatible with the other filters."

addedInput schema / properties / filters / properties / min_height / description
```
Added value: +"Minimum image height in pixels. Omit for no height floor."
```

addedInput schema / properties / filters / properties / min_width / description

Added value: +"Minimum image width in pixels. Defaults to 1000 when omitted; null removes the width floor."

addedInput schema / properties / filters / properties / monochrome / description

Added value: +"Filter by monochrome status. Omit to allow both monochrome and color images."

addedInput schema / properties / filters / properties / nsfw_max / description

Added value: +"Maximum recorded NSFW score from 0 to 1; lower is stricter. Defaults to 0.2. Assets without a score can still match."

addedInput schema / properties / filters / properties / orientation / description

Added value: +"Required image orientation. Omit to allow landscape, portrait, and square images."

addedInput schema / properties / filters / properties / source / description

Added value: +"Allowed source IDs from result.source, for example [\"flickr\"]. Omit to search all sources."

addedInput schema / properties / filters / properties / year_max / description

Added value: +"Latest year, inclusive. Use with year_min to bound the date range."

addedInput schema / properties / filters / properties / year_min / description

Added value: +"Earliest year, inclusive. Images without matching year metadata are excluded."

addedInput schema / properties / k / description

Added value: +"Maximum number of final results (1–100, default 10). Fewer results may match the filters; this is separate from candidate-pool sizes."

addedInput schema / properties / mode / description

Added value: +"Ranking mode: visual selects the default relevance reranker; none skips reranking and returns retrieval order. Default visual."

addedInput schema / properties / query / description

Added value: +"Describe the image, subject, scene, or mood to find (1–1000 characters), for example: misty mountains at sunrise."

addedInput schema / properties / rerank_k / description

Added value: +"Maximum candidates to rerank (1–500, default 60); null reranks the whole candidate pool. Ignored when mode is none. Independent of final result count k."

addedInput schema / properties / rerank_text / description

Added value: +"Legacy reranker text toggle (default true). The current service always uses text; leave true."

First observedSep 22, 2026

TDQS

A4.9/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations are minimal (readOnlyHint=false, destructiveHint=false, openWorldHint=true, idempotentHint=false), so the description carries the burden. It discloses cost ('Each successful call charges $0.02 and records usage'), auth requirements, rate limits, the four defaulted filters with null opt-outs, and side-effect scope ('Does not modify images or publish anything'). This far exceeds what annotations provide.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Dense but every sentence earns its place: scope, inputs, exclusions, sibling routing, return shape, defaults, auth, cost, and safety. High-value constraints like 'unknown NSFW scores pass' and 'repeated calls can charge again' are included without bloat.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

A complex 12-parameter tool with a nested filters object and no output schema; the description covers the return format (query_id, ranked results with asset_id, file and thumbnail URLs, dimensions, rights), cost model, default behaviors, and exclusions. Nothing an agent needs to invoke it correctly is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% so the baseline is 3, but the description adds genuinely useful semantic context: defaults for the four filter values and how to opt out ('these four defaulted filters accept null to opt out'), that unknown NSFW scores pass, and that k is separate from candidate-pool sizes. This clarifies relationships the schema does not.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource: 'Search the Voyage index of 1,852,370 captioned images using joint image-and-caption embeddings, hybrid retrieval and caption text reranking.' It names input modalities (text, image_base64, image_url, image plus text) and explicitly excludes icon-pack search, distinguishing itself from siblings.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly routes to alternatives: 'Use find_similar_images when you already have a Lightdrift asset_id, or get_image to retrieve one known image without a search.' It also tells the agent what to supply ('Supply query text or one image'), how to phrase queries, and what is unsupported ('icon-pack search is not supported').

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Lightdrift Images MCP tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources

Need Help?