Getting started

1Install

Download the installer for your platform from the download section or GitHub Releases ↗.

• macOS: Open the .dmg, drag Condit to Applications. First launch: right-click → Open to bypass Gatekeeper.
• Windows: Run the .exeinstaller. Click “More info” → Run anyway on the SmartScreen prompt.

2Connect a database

1. Click + New connection in the left sidebar.
2. Choose your database type from the dropdown.
3. Enter the connection details (host, port, user, password, database name).
4. Click Test to verify, then Save.

Connection credentials are encrypted and stored locally — never sent anywhere.

3Ask a question

Select a connection, type in the chat bar at the bottom, and press Enter.

Condit writes the SQL, runs it against your database, and returns the results. You can copy the SQL, edit it in the editor, or ask a follow-up question.

4SQL editor

Prefer writing SQL directly? Use the editor tab next to the chat. Press Cmd+Enter to run. Results appear in the panel below with column types, row count, and elapsed time. Right-click any result header or cell for copy options.

5LLM setup

Go to Settings → LLM → Add profile. Condit works with any of these providers:

OpenAI

gpt-4o, gpt-4-turbo, gpt-3.5-turbo

Anthropic

claude-3-5-sonnet, claude-3-haiku

Gemini

gemini-1.5-pro, gemini-flash

Groq

llama-3, mixtral — very fast

Ollama

Any local model — fully offline

For Ollama: set provider to Ollama, base URL to http://localhost:11434, and pick any pulled model. Fully local, zero API costs.

If you don’t have an LLM of your own, Condit AI is a hosted model tuned for SQL that’s ready instantly — no API key or local install, $12/month. Turn it on from Settings → Condit AI.

6Semantic layer (optional)

Right-click any connection in the Navigator → Semantic Layer. Add plain-English descriptions for its tables and columns. The AI injects these automatically so you get better SQL without re-explaining your schema every session.

Use the AI Assist button to have the LLM suggest descriptions based on the column names and sample data.

7Data privacy

Settings → Data Privacy shows exactly what goes to your LLM provider: table and column names and relationships, always; up to 50 rows of query results to generate the plain-English answer; and never your database credentials or connection strings.

Turn on Schema-only mode to stop query results from ever being sent — SQL generation still uses the LLM, but the plain-English interpreter and result verifier are skipped.

8Share config with a team

Settings → Team Config exports your scope presets, semantic layers, and LLM profile names as a JSON bundle. A teammate can import it to pick up the same setup without retyping API keys.

Keyboard shortcuts

Updates

Condit checks for updates automatically while it’s open. Current version and a manual check live in the status bar at the bottom of the window.