Schema-aware editing
Fast, local completion for tables, columns, joins, CTEs, packages, and procedures.
Oracle-first desktop workspace
Your SQL client,
with an agent inside.
SQLLM combines the database tools developers already need with AI that understands schemas, queries, errors, execution plans, and PL/SQL relationships.
Working prototype · Windows · Oracle 11g compatible
orders.sqlRun
SELECT
c.segment,
COUNT(o.order_id) AS orders,
SUM(o.total_amount) AS revenue
FROM orders o
JOIN customers c
ON c.customer_id = o.customer_id
WHERE o.created_at >= :start_date
GROUP BY c.segment;Results3 rows · 18 ms
| SEGMENT | ORDERS | REVENUE |
|---|---|---|
| Enterprise | 1,248 | ₩84.2M |
| Growth | 973 | ₩51.7M |
| Starter | 2,104 | ₩19.8M |
Beyond Text-to-SQL
Context is what turns a model into a database teammate.
01
A prompt box
Text-to-SQL starts with a question and ends with generated syntax. The developer still has to reconstruct the database context.
02
A working environment
SQLLM brings the schema, server version, active SQL, Oracle error, execution plan, and PL/SQL relationships into the conversation.
One workspace
Query. Understand. Debug. Operate.
AI with real context
Generate, explain, fix, and optimize using the relevant schema, error, and execution plan.
PL/SQL intelligence
Trace calls, CRUD, impact, long cursors, and data flow across legacy packages.
Agent-ready MCP
Give Claude and compatible agents read-only access to allow-listed metadata and saved analysis.
Database operations
Inspect plans, sessions, locks, transactions, and paged results without leaving the workspace.
Local or frontier models
Choose Ollama, LM Studio, vLLM, compatible endpoints, or Anthropic for each environment.
Human in control
AI proposes.
You review and run.
01AI output never executes automatically.
02The MCP server exposes no arbitrary SQL execution tool.
03Query results and passwords are not sent to model providers.
04Read-only profiles and destructive-query checks add guardrails.
Building in public
Database work deserves an agent that understands the whole system.
SQLLM is an active prototype, beginning with the Oracle environments that generic AI coding tools often overlook.