Most AI database tools have a dirty little secret. Your queries, your schema, your row data, all of it passes through someone else’s server before anything useful comes back. That middleman problem has quietly made a lot of privacy-conscious developers and analysts walk away from the category entirely. It’s a real shame, because the underlying idea, asking your database questions in plain English, is genuinely useful.
dbFlow.ai takes a different position. It’s a native desktop app for macOS, Windows, and Linux that connects directly from your machine to whichever LLM provider you choose. No dbFlow server sits in the middle. Your credentials live in your OS keychain. And if you enable local inference with Ollama, nothing about your data leaves your machine at all. That last part is what makes it worth paying attention to.
Jake Sherwood, the developer behind it, puts it plainly: point Little Snitch at the app and watch. You’ll see your database connection, your chosen LLM provider, a license check, and an update ping. That’s the full list. For anyone working with sensitive data, that kind of verifiability matters more than any privacy policy ever could.
Key Features of dbFlow.ai
Natural Language to SQL, No Expertise Required
You type a question in plain English. dbFlow generates the correct SQL, runs it against your database, and explains the results. It supports PostgreSQL, MySQL, SQLite, and MongoDB. The agent also maintains session context, so follow-up questions build on what came before rather than starting from scratch each time.
Fully Local Inference with Ollama
This is the standout feature. Enable Local Inference in settings, point dbFlow at your localhost or a GPU box on your LAN, and the entire workflow runs on your own hardware. Prompts, schema details, and query results never touch an external server. One-click model installs are available for Qwen 3.5 and Llama 4 Scout, sized for laptops through power-user rigs. Models warm up in VRAM so queries feel fast after the first one.
Self-Correcting Agent
Generated SQL doesn’t always come out right on the first try. dbFlow catches errors automatically, fixes the query, and refines its approach over time within a session. You don’t have to intervene or re-explain the question. It just sorts itself out and returns the correct result.
Charts, Dashboards, and Read-Only Safety
Any result set can be turned into a bar chart, pie chart, or dashboard with a single command. Shared dashboards and a shared query library are available on Team and Enterprise plans. Critically, dbFlow is read-only by design. It only generates SELECT queries. There are no dropped tables, no accidental writes, no unexpected side effects.
How a Healthcare Analyst Used It at 7am Without Sending a Single Row to the Cloud
Priya Nair is a data analyst at a mid-sized healthcare clinic. Every Monday morning at 7am, before the rest of the team logs on, she needs to pull the previous week’s appointment no-show rates broken down by department. Her production database holds patient records, which means cloud AI tools are completely off the table under her organization’s data policy.
She downloaded dbFlow, connected it to the clinic’s PostgreSQL instance, and configured Local Inference using a Qwen 3.5 model running on her workstation. At 7:03am that Monday, she typed: “Show me no-show rates by department for the past seven days.” Thirty seconds later she had a clean table and a bar chart ready to screenshot into her weekly Slack update.
No data left the building. No one reviewed her query on a remote server. She verified this herself by checking the app’s network activity, exactly as Jake suggests. The whole process took under four minutes, compared to the twenty-plus minutes she’d previously spent writing and debugging the SQL herself. That outcome repeated the following Monday, and the Monday after that.
How dbFlow Actually Works, Step by Step
Step 1: Download and Install
Grab the native desktop app for your platform. macOS, Windows, and Linux are all supported, with automatic updates built in.
Step 2: Connect Your Database
Enter your credentials. They go straight into your OS keychain and are never transmitted to dbFlow’s servers. You can connect multiple databases and switch between them in a single session.
Step 3: Choose Your Model
Pick a cloud model (GPT, Claude, Gemini, DeepSeek) using bundled credits or your own API key. Or flip on Local Inference in Settings and point the app at an Ollama endpoint on localhost or your private LAN.
Step 4: Ask Questions and Refine
Type your question in plain English. Review the generated SQL if you want to. Iterate in natural language. Turn results into charts or dashboards with a single follow-up command.
dbFlow.ai Pricing
dbFlow offers a free tier for getting started. Paid plans are billed monthly or annually:
- Pro: $19/month (or $190/year). 15 databases, 200 AI queries per day, 6 AI models, full visualizations, BYOK support.
- Team: Unlimited databases, 500 queries per day per user, shared dashboards and query library, priority support. Contact for pricing.
- Enterprise: Unlimited queries and tokens, self-hosted and air-gapped options, SSO/SAML, custom model support, dedicated support.
All paid plans include Local Inference with Ollama, and local model queries do not count against your AI credit limits. A 30-day money-back guarantee applies to paid plans.
Who Should Try dbFlow.ai
If you work with a production database and don’t have a dedicated analyst on call, dbFlow fills that gap quickly. It’s especially well suited for teams in healthcare, legal, and finance where sending data to cloud AI tools isn’t an option. E-commerce operators tracking refund rates or subscription renewals will also find it genuinely practical. Download it free at dbflow.ai and run your first query in under ten minutes.
