Your agent has the rows. Give it the relationships.

An agent that can only SELECT sees a table at a time. The questions that decide things — who else does this supplier touch, what is five hops from this account, which of these are the same company — are traversals. Declare a graph over the tables you already have and your agent can walk them: no graph store to load, nothing to sync, writes still ordinary SQL with transactions.

Your tables stay the truth.

CREATE GRAPH builds a compact adjacency artifact pinned at a specific table version and serves traversals from it. It rebuilds itself when the edge table drains or compacts — and when you need the newest writes before that happens, a 'strong' read merges them into the traversal at query time.

  nodes/edges tables ──CREATE GRAPH──►  adjacency artifact @ version V
        │                                      │
        │  writes (SQL INSERT/DELETE)          ├── default read: artifact as-is
        ▼                                      │
  write-ahead log ────────'strong' read────────┴── artifact ∪ un-drained writes

Reach it however you already work.

Declare it once. Query it anywhere.

-- ordinary tables; the edge table needs its own PK
CREATE TABLE person (id INT PRIMARY KEY, name VARCHAR);
CREATE TABLE knows  (edge_id INT PRIMARY KEY, src INT, dst INT);

-- declare the graph over them
CREATE GRAPH social NODES (person KEY id) EDGES (knows SRC src DST dst);

-- traverse
SELECT * FROM graph_khop('social', 1, 2);            -- (node, hop_distance)
SELECT * FROM graph_neighbors('social', 2, 'both');  -- (neighbor)
SELECT * FROM graph_shortest_path('social', 1, 7);   -- (step, node)

-- whole-graph analytics
SELECT * FROM graph_pagerank('social');              -- (node, rank)
SELECT * FROM graph_components('social');            -- (node, component)

Traversals and whole-graph analytics.

Traverse in a subquery. Aggregate outside it.

Because the graph lives in the same engine as the rows, a traversal is just another relation. No round trip to a second database, no stitching two result sets together in application code.

-- total deal value across an account's 5-hop corporate family
SELECT SUM(amount) FROM deals
WHERE account_id IN (SELECT node FROM graph_khop('accts', 1, 5));

Four dimensions over one copy of the data.