ShrikeDB Cloud

Create an account at shrikedb.com/signup. Registration requires a workspace name, email address, and password. A credit card is not required.

Create a workspace

  1. Enter your workspace name, email, and password.

  2. Open the verification email from no-reply@shrikedb.com.

  3. Enter its six-digit code on the verification screen.

  4. Sign in at the console. Your verified account receives its own workspace.

Verification is required before the service provisions a database. Retrying sign-in opens the same workspace. Account identity remains bound to the identity-provider subject, rather than to an editable email address.

The workspace has six destinations:

Destination Use

Data

Choose or drop a file, then inspect its fields.

Schema

Review the sampled types and map rows to nodes or relationships.

Query

Run read-only Cypher or SPARQL and export returned rows to CSV.

Chat

Ask questions in plain language and inspect the query behind each answer.

Vectors

Load embeddings, search a named index, and open the matching graph query.

API endpoints

Publish an endpoint and create a private application key.

Import and query data

Free workspaces accept uploaded CSV, TSV, JSONL, Turtle, RDF/XML, and OWL files up to 2 MiB and 10,000 lines. JSONL supports inspection; convert it to CSV or TSV to load. Convert XLSX files to CSV or TSV before uploading to a free workspace.

Inspect the file, review the fields, choose the mapping, and select Load dataset. After loading, select Query data to populate a query for the dataset. The query editor retains separate Cypher and SPARQL drafts within the current browser session. Press Ctrl+Enter, or Command+Enter on macOS, to execute.

The console returns up to 1,000 rows. Export CSV downloads the returned rows. Recent queries remain available during the current session.

Chat with your data

Open Chat after loading a dataset. Ask a specific question such as “How many nodes are in this workspace?” or “Show me a few records.” The assistant prepares a read-only Cypher or SPARQL query from your workspace schema, runs it in your own database, and summarizes the returned rows. Open View query under an answer to see the executed statement and row count. Select Open in query editor to inspect or refine it yourself.

You can ask follow-up questions in the same conversation. New chat clears the conversation in this browser session. The chat does not write to your database. Questions, a short conversation history, schema names, and selected query results are sent to OpenAI to generate answers. Do not put secrets into questions. The OpenAI credential remains on the ShrikeDB server; it is never placed in the page or returned to the browser.

Chat executes at most 30 returned rows per question. For a complete export, use the Query editor or your application API. An answer describes the query result available at the time of the request; open the query to check its scope. The shared cloud service permits 60 chat questions per workspace per hour. If the hourly limit is reached, wait before sending another question.

The initial free plan has a 10,000-record graph import allowance and a 64 MiB encoded import budget. The service reserves the budget before writing. A failed or uncertain write retains its reservation; ordinary completed in-process retries use the existing load receipt. Three active API keys and three active published endpoints are available per workspace. Hosted capacity is bounded by the shared service’s configured resources.

Connect an application

Publish an endpoint from API endpoints. Choose the accepted query languages and copy its URL. Create an API key and copy the secret immediately; the console reveals it once. Store it in your application’s secret configuration.

For example, after setting SHRIKEDB_ENDPOINT to the copied URL and SHRIKEDB_API_KEY to your private key:

curl "$SHRIKEDB_ENDPOINT" \
  -H "Content-Type: application/json" \
  -H "x-shrikedb-api-key: $SHRIKEDB_API_KEY" \
  --data '{"mode":"cypher","query":"MATCH (n) RETURN count(n) AS total","limit":10}'

The endpoint is bound to your workspace. A client-supplied database header does not change that binding. Revoke unused keys and endpoints from the console. The free API surface supports inspected dataset uploads, direct read-only queries, and published query endpoints.

Keys and permissions

An application key authenticates as the user who created it and inherits that user’s current workspace permissions. A published endpoint restricts the accepted query languages; it does not restrict the key to that endpoint. The same key can use the direct read-only query API and other operations permitted by the user’s role and hosted plan. Use a dedicated application identity when your deployment provides one.

Send the key in the x-shrikedb-api-key header over HTTPS. Keep it in server-side secret configuration; exclude it from browser code, URLs, source control, and logs. Key listings display metadata and a short identification prefix, never the secret or its stored hash. Revoke a key if it is exposed, then create a replacement. Keys created in the console remain valid until revoked. Revocation takes effect on the next request; it does not cancel work already admitted.

Direct queries

Use the same key with the direct query API when you do not need a published endpoint:

curl https://shrikedb.com/api/v1/query/direct \
  -H "Content-Type: application/json" \
  -H "x-shrikedb-api-key: $SHRIKEDB_API_KEY" \
  --data '{"mode":"cypher","query":"MATCH (n) RETURN count(n) AS total","limit":10}'

For SPARQL, set mode to sparql and supply a read-only query such as SELECT (COUNT(?s) AS ?total) WHERE { ?s ?p ?o }. Both query routes return JSON with fields, rows, returned_rows, truncated, and elapsed_ms. The limit controls returned rows, up to 1,000; it does not limit the work needed to compute an aggregate. elapsed_ms measures server-side query processing after authentication, rather than network round-trip time.

Reuse an HTTP connection for repeated calls. Prefer aggregates and targeted projections when an application needs a summary. The shared free service admits two concurrent data operations; HTTP 429 indicates that its available request slots are occupied. Retry with a short increasing delay. HTTP 401 indicates an absent or invalid credential; HTTP 403 indicates insufficient permissions, a suspended workspace, a disabled endpoint language, or an attempted write query. An inaccessible or revoked published endpoint returns HTTP 404.

Recover and protect an account

Select Forgot password? on the sign-in screen, or open password recovery. Enter your email, then use the delivered reset code to set a new password. A successful reset revokes previous application and identity-provider sessions. Application API keys remain separate credentials.

For two-step verification, select Account, choose Set up two-step verification, add the displayed setup key to an authenticator app, and enter its current six-digit code. Subsequent sign-ins request an authenticator code after the password. Store the authenticator’s recovery material securely. Password recovery preserves the second-factor requirement.

Account emails come from no-reply@shrikedb.com. Incoming replies to that address are accepted and discarded.

Platform administration

The deployment owner signs in at Administrator sign-in with the platform owner key supplied through the deployment’s provisioning procedure. The platform page lists tenant status and import-budget usage, and supports suspension or restoration with a recorded reason. Tenant administrators use their own workspace credentials; platform administration requires the owner identity.

The administrator page keeps the supplied owner key in memory and clears the input after successful authentication. Closing the page removes that browser session’s key. Protect the owner key as a deployment secret.