OriginChainDB documentation
OriginChainDB is a managed database that answers five kinds of question against the same data: SQL for filters and aggregates, vector for similarity, full-text for keywords, graph for relationships, and natural language for everything else.
All five share one endpoint, one bearer token, one row store. A row written once is visible to all of them. New to the database? Hit the quickstart below. Already shipping? Jump to the per-shape reference pages.
One row store, five kinds of question.
A row is written once. Every surface below reads that same row through the same endpoint and the same bearer token. There is no copy to keep in sync, no second store to provision, and no window where one surface can see a write the others cannot.
Which rows match?
What is it similar to?
Which mention this?
How are these linked?
Just ask in English.
Not sure which to reach for? Choosing a query shape compares them side by side on the same data.
Five minutes from signup to your first SQL, vector, full-text, and graph query - with cURL, Python, TypeScript, and Go side by side.
curl -X POST "https://$OC_HOST/v1/tenants/$OC_TENANT/rows/shop.orders" \
-H "Authorization: Bearer $OC_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"id": "01JTRX9KQ3YH8K2WMX0F5JZAB7",
"customer": "01JTRX1H4Q9P0N2WMX0F5JZ001",
"amount_cents": 12950,
"status": "paid",
"notes": "rush delivery, signed by recipient",
"placed_ms": 1714478049000
}'curl -X POST "https://$OC_HOST/v1/tenants/$OC_TENANT/sql" \
-H "Authorization: Bearer $OC_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"sql": "SELECT customer, SUM(amount_cents) AS total FROM shop.orders WHERE status = '\''paid'\'' GROUP BY customer LIMIT 10"
}'Both calls are lifted from the quickstart, where the same steps are shown in cURL, Python, TypeScript and Go. Vector, full-text and graph reads against that row follow the same pattern.
Pick the right one for your question.
Each shape gets its own reference page with cURL, Python, TypeScript, and Go examples. Not sure which to use? Choosing a query shape has a side-by-side comparison.
SELECT with WHERE, GROUP BY, aggregates (COUNT/SUM/AVG/MIN/MAX), and INNER / LEFT / RIGHT / FULL OUTER JOIN across up to 32 tables. Plus EXPLAIN.
Top-k nearest-neighbor over embeddings. HNSW by default; IVF and IVF-PQ for 10M+ corpora. Cosine / dot / L2 / Manhattan, optional metadata filter.
BM25 ranked, boolean AND, or exact-phrase. 18-language stemming and 9-language dictionary lemmatization. Optional diacritic folding and stop-words.
19 algorithms - neighbors, BFS, shortest path, Dijkstra, PageRank, Louvain, Node2Vec, GraphSAGE. Edges piggy-back on row columns; declare a relation on the schema and the engine maintains them.
Send a sentence in English, get rows back. The compiler caches plans by question hash so repeats are fast.
Or by topic.
Choosing a query shape, SDKs, the CLI, the dashboard, deployment, operations.
Decision guide. If your query looks like X, reach for Y - side-by-side per-shape comparison.
Python, TypeScript, and Go clients. Auth, idempotency keys and typed responses handled for you. Python also retries transient failures; TypeScript and Go leave the retry to you.
Step-by-step guides for inserts, schemas, queries, backups, and restoration in the managed console.
Region pinning, configuration choice, dedicated single-tenant compute, sync follower, managed TLS, managed DNS.
Backups, point-in-time recovery, metrics, audit log, failover.
Auth, errors, limits, fixes.
The platform reference - things you need once, then come back to when something breaks.
Bearer tokens - how to get one, store it, rotate it, and what to do if it leaks.
Per-token limits (1,000/min default), 429 handling, and Retry-After.
Catalog of error codes - 4xx and 5xx - with cause, recovery, and SDK retry policy.
Common problems organized by symptom, with the fix alongside.
Every endpoint - schemas, rows, sql, vector, fts, graph, ask, migrations, replication, health. Bearer auth, idempotency keys, rate-limit headers.
@originchain/mcp-server - drop into claude_desktop_config.json and your AI agent can ask, SQL, vector-search, FTS, and list schemas on your tenant.