The OpenGraph Image MCP server exposes five operations. Together they cover graph construction, discovery, inspection, and natural-language querying.
| Tool | Input | Use it when… |
|---|---|---|
build_graph |
folder_path |
A source folder does not have a graph yet, or needs to be rebuilt |
query_graph |
question, graph_path |
You need a grounded answer from the graph |
get_image_entities |
image_filename, graph_path |
You want every entity and edge connected to one image |
list_images |
graph_path |
You need to see which source images are represented |
graph_summary |
graph_path |
You want high-level graph statistics before querying |
build_graph
Builds a graph from every supported image in a folder.
build_graph(folder_path="./photos")
Use the returned graph path for later inspection and query calls.
query_graph
Answers a natural-language question using an existing graph.
query_graph(
question="Which images show a dog near water?",
graph_path="./photos/opengraph-out/graph.json"
)
get_image_entities
Returns the nodes and edges associated with a specific source image.
get_image_entities(
image_filename="park-03.jpg",
graph_path="./photos/opengraph-out/graph.json"
)
list_images
Lists every image currently represented in a graph.
list_images(graph_path="./photos/opengraph-out/graph.json")
graph_summary
Returns a compact overview of the graph and is a useful first call when an agent inherits an unfamiliar workspace.
graph_summary(graph_path="./photos/opengraph-out/graph.json")
Agent pattern: Summarize first, query second, then inspect individual sources only when the answer needs deeper verification.