Governed data work, with AI assist
CKN Intelligent is the governed workbench for your databases, data pipelines and analysis, with AI help for drafting SQL. Every connection, grant and statement is approved, bounded and recorded, so teams move faster without losing control.
The health of all your data
Unreachable connections, access about to expire and overdue backups, visible before they hurt.
Every connection registered and probed
Know which databases are production, who owns them and what their last probe said.
What CKN Intelligent does
Governed access to your databases
Register a database once, then give people time-limited grants instead of standing passwords. Every statement is checked before it runs and every session is recorded. Explain plans, ER diagrams, schema compare and exports to JSON, CSV, JSONL or XLSX are built in.
- Read, write and schema grants, approved by a second person
- Statements classified by parsing, not by their first keyword
- Tenant filter, column masking and row limits applied for you
- PostgreSQL in full; MySQL, MariaDB and SQL Server through the console
AI help for writing SQL
Describe what you need in plain words and get a draft statement in the editor. The AI sees table and column descriptions only, never your rows, and nothing it writes runs until a person runs it.
- The draft is pre-checked against your grant before you run it
- Masked columns are flagged and literal values are stripped before anything is sent
- Off by default; you choose the provider (adapters for Claude, ChatGPT and Gemini)
- Every call is audited with token counts, without keeping the request text
Bulk changes with preview and undo
Mass corrections follow five stages that cannot be skipped: plan, preview, approve, execute and verify. If the data shifts after approval, the run stops and asks again.
- Every write needs a predicate; one without it is refused at plan time
- Before-images are captured ahead of every write batch
- Reversal inside a declared window, skipping rows changed since
- Actual effect compared with the approved estimate, with any variance reported
Backups you can verify
Backup policies run on a schedule triggered through CKN Integration, and artifacts are hashed and encrypted. An artifact that was never verified is reported as unverified, not assumed good. Bulk changes, backup and restore run on PostgreSQL.
- Trial restore to a scratch target when the policy asks for it
- Restore needs a second approver and an explicit acknowledgement that downstream systems are not rewound
- A production artifact cannot go into a lower environment without a declared masking or subsetting step
- An alert is raised when a scheduled backup has not run
Data pipelines with lineage
Build versioned ETL or ELT flows that copy data from registered sources into a warehouse. Pipelines read under read-only grants and never write to a source. The SQL for each transform is generated from declared steps, not typed in.
- Full, incremental or change-window extracts with checkpoints
- The checkpoint moves forward only after the load is verified
- Rejected rows are kept with a reason, never silently dropped
- Lineage traced from the source object to the warehouse model
Analysis workspaces and a model registry
Analysts work in workspaces with their own data limits and compute ceilings. Notebooks run in a sandbox with no database credentials and no outbound network.
- Datasets are masked when built, so hidden columns are not in the file at all
- Experiment runs record dataset version, method, parameters and seed
- A model cannot be registered without its training data and method version
- Model output is a proposal, never an automatic action
Built and tested at this scale
These figures are counted from the product and its test suite. They are not speed benchmarks.
Figures come from CKN's own test suites and benchmarks. Real results depend on each system's hardware and data volume.
What your business gets
Know who touched what
Every database touch records the person, the purpose, the grant, the statement and the effect. There are no standing passwords, and grants expire on their own.
Lower risk when data must change
Preview before a bulk write, undo inside a window, and backups reported as verified or unverified. Mistakes stay smaller and are easier to recover from.
Faster answers without open access
Analysts and support staff get masked data, saved queries, exports and AI-drafted SQL within their own limits. No shared database login is needed, and it works with the databases you already run.
Run a statement under an approved grant
Who relies on it
- CKN Integration, which calls it for the operator and triggers scheduled backups, pipelines and clean-up jobs
- Operators who sign in directly through CKN Identity while CKN Integration is unavailable
- Support and operations teams looking after the databases behind the CKN business services
- Data and analytics teams building pipelines, datasets and models
POST https://intelligent.example.com/access/session/execute-statement
Authorization: Bearer <integration-access-token>
X-CKN-Delegation-Assertion: <delegation-assertion>
X-CKN-Operation-Code: INTELLIGENT.ACCESS.SESSION.EXECUTE_STATEMENT
X-CKN-Grant-Id: <grant-id>
X-CKN-Purpose: SUPPORT
X-Correlation-Id: <uuid>
Idempotency-Key: <uuid>
{"sessionId": "<session-id>", "statement": "SELECT id, tenant_id FROM orders"}
HTTP/1.1 200 OK
{"success": true, "data": {"statementClass": "READ", "tenantScope": "SINGLE_TENANT",
"scopedObjects": ["orders"], "rows": [{"id": "...", "tenant_id": "..."}]}}