The Desktop Powerhouse
Octofy Pro leads with intelligent SQL assistance: turn natural language into SQL for the connection you have open, improve the script already in your editor, or narrow in on runtime errors—with optional context from conversation history, pinned objects, and verified value mappings. Use the Built-In Agent for a local, schema-indexed pipeline (knowledge-base-first discovery, vector indexes, validation loops, and thumbs-up examples saved on your machine), or wire in the Octofy Agent enterprise API for hosted discovery, streaming SQL generation, schema administration, and remote few-shots. Layer business meaning on top: a semantic layer of measures, dimensions, joins and governance predicates that compiles a structured request into physical SQL; data groups whose rules and precomputed Q&As steer discovery; tool functions the agent can call—locally or over MCP—before it writes a filter; and Learn from Past Projects, which mines business rules, value mappings and semantic models out of your existing SQL, SAS, R and Python scripts. It connects natively to SQL Server, PostgreSQL, MySQL (including MariaDB) and Oracle, with any other engine reachable through ODBC. When you need to interpret results, load from a table or view, an arbitrary SELECT, or the preview pane; Octofy classifies columns and drives nine analysis modes—counts and sums by category with date grouping (month, quarter, year, ISO week), configurable percentiles, distribution pivots for counts and sums (including row-percent views), and scatter bubbles sized by count or sum—shown as bar, stacked, time-series, and bubble charts, with filters, sortable summary pivots, drill-down to raw rows, chart copy, quick charts, and column value frequency. Prefer a canvas? A visual query designer still lets you compose joins and inspect generated SQL next to the editor. The interface is localized into Deutsch, Español, Français, 日本語, 한국어, 简体中文 and 繁體中文 alongside English, and follows your Windows display language.
Feature Carousel
Natural language to SQL, script optimization, and error-aware fixes with local schema indexes or the Octofy Agent API.
Visual query designer preview
When you compose joins on the canvas, choices map to SQL like this—pick a join type to preview.
Only matched records
SELECT c.CompanyName, o.OrderDate
FROM Customers c
INNER JOIN Orders o ON c.CustomerID = o.CustomerID
WHERE o.OrderDate >= DATEADD(day, -30, GETDATE());Knowledge-base-first discovery with a deterministic exact-question match, few-shot retrieval, live column hydration with pruning, multi-attempt validation (including result-set metadata), and a thumbs-up flow that saves reviewed question/SQL pairs on your machine—the canonical wording the agent recognises instantly next time.
Describe the business meaning of your data as measures, dimensions, joins and governance predicates. The agent emits a structured request that compiles to physical SQL against the active model, with a compile preview before you save. Models accumulate across runs, and a definition that conflicts with a stored one is escalated for review rather than silently overwritten.
Group objects the way the business talks about them—orders, customers, invoices—and attach business rules, metrics, keywords and reviewed question/SQL pairs. Groups steer discovery at query time; an approved pair that matches closely is reused as the answer, and looser matches become few-shot context. AI can draft pairs and auto-approve the ones that generate valid SQL.
Give the agent callable functions so it can look something up before writing the filter. A tool runs either as a local sub-agent with its own model and system prompt, or over MCP against a remote server. Arguments are validated against the declared parameter schema, and whatever the tool returns feeds the query analysis.
Point Octofy at a folder of legacy SQL, SAS, R, Python, Java, C or C# scripts. It extracts business rules with file-and-line evidence, value mappings, few-shot examples, precomputed Q&As, skills and semantic models—with a quarantine ledger for anything it refuses. Runs are resumable, and re-running a folder adds no duplicates.
Let the configured LLM write the documentation: describe the selected table or view, batch-describe the whole library, or fill in only the gaps. Descriptions are written into the library's markdown files—the source of truth—and the object's vector entry is refreshed so search reflects the new text.
Work against your databases on your own machine without shipping row data to a cloud LLM or public endpoints. The built-in agent generates SQL from schema metadata only—table and column names, types, and relationships—keeping sensitive cell values local.
Optional remote agent with discovery, SSE SQL generation, schema sync administration, and hosted few-shot storage when your organization runs the API.
Start from the object tree (Analyze), query results, or preview Chart—same engine whether you load a full table or a limited row set. Variable roles (category, distribution, sum) respect column typing and cardinality limits; the filter panel lets you include or exclude values and date ranges, then reapplies every aggregation on the filtered subset. Summary grids pair with a raw-data grid for drill-down, exclude-blanks, clipboard copy, and preview export to CSV or Excel.
Compose joins on a canvas, adjust relationships visually, and review the generated SQL alongside the text editor when you want a graphical starting point.
Discuss the previewed dataset with any configured LLM provider right inside the Data Preview window. The assistant sees the query, the visible columns and their types, and a bounded sample of rows, then streams markdown answers — summaries, comparisons, and pattern explanations — while an explicit consent setting controls when row data leaves your machine.
The application interface is localized for Deutsch, Español, Français, 日本語, 한국어, 简体中文 and 繁體中文 alongside English, and follows your Windows display language—so the UI reads in the language your team already works in.