Desktop

The workbench

One desktop app for the whole job: bring data in, query it, and hand the keys to the coding agent you already use.

Speak to your data, your way

Natural language, SQL, or the database's own language. Whichever you write, SnoutData runs what the database actually executes.

Ask in plain English

Tell the assistant what you want. It reads your live schema and writes a query that runs.

SQL, on anything

Write SQL against MongoDB or Pinecone and SnoutData transcribes it to a native pipeline or vector query, offline for the common shapes.

Native, full power

Or write MongoDB's aggregation pipeline directly, with schema-aware completion and the same results grid.

Connect to the databases you already use

One IDE, one assistant, across relational, document, and vector backends. More engines land regularly.

PostgreSQLRelational

MySQLRelational

Amazon AuroraRelational

MariaDBRelational

SQL ServerRelational

OracleRelational

IBM Db2Relational

SQLiteRelational

DuckDBRelational

SnowflakeRelational

Amazon RedshiftRelational

SAP HANARelational

MongoDBDocument

PineconeVector

A real data workbench, with a brain

Everything a professional database client gives you, plus a way in for your data and an agent that can drive it.

Every database, one IDE

Relational, document and vector databases in one editor, one results grid, one assistant. Fourteen engines, each with a native driver.

Schema-aware AI

It reads your live tables, columns and foreign keys, so it drafts queries that fit your database instead of generic guesses.

Tuned, not token-hungry

A smaller, well-tuned model gets exactly the tables your question touches. No crawl of your whole database, no wasted tokens.

A real data workbench

Page through millions of rows, edit cells inline, and manage tables, collections and indexes with a preview that runs on confirm.

Local-first and private

A native desktop app, not a browser tab. Passwords and SSH keys are encrypted in your OS keychain, never in config files.

Bring your own model API

On Plus and Pro, point the assistant at your own model. Your key and prompts go straight to the provider and never touch our servers.

A way in for your data

Land a file, a PDF, a web page or another database into a table, a collection or a vector index, on a schedule, with a ledger of every run.

Your coding agent, inside the app

Claude Code, Codex and opencode run in a dock with your databases already connected, or point any MCP client at it from outside. The agent never receives your credentials.

Safe on production

Production connections are flagged, destructive statements ask before they run, and every query is logged.

The assistant

It reads your schema before it answers

SnoutData ranks the tables your question touches, follows their foreign keys, and grounds the model in your real columns and types. So the query it writes runs.

Draft SQL or a MongoDB pipeline from a plain-English question
Explain, refactor, or optimize a query you already have
Ask: quick, one-shot answers to a single question
Agentic: extended thinking for deeper, multi-step work
It runs only when you approve, and every plan can keep an audit log of every action

Try it free

Which products have no orders in the last 90 days?

reading schema: products, orders

select p.name
from products p
left join orders o
  on o.product_id = p.id
  and o.created_at > now() - interval '90 days'
where o.id is null;

See it in action

A real desktop app for working with data, with an assistant that understands it.

Ask in plain English

Describe the report, get the SQL

It reads your live schema, writes runnable SQL, and explains why. One click puts it in the editor.

Tab to complete, our own model

Snout Complete finishes the line for you

Start typing, press Tab. This is our own FIM model, snout-fim-1.0, served so the suggestion fits your actual schema.

It learns how you work

Sharper the more you use it

It learns your patterns, your formatting, and the columns and filters you reach for, then folds that into its suggestions.

Agent mode

Let the assistant run the query

Expand to a full window and let it run queries, read the results and keep going. You choose the model and how much freedom it gets.

Stuck on an error

Paste a failing query, get the fix

Hand it the error and the SQL. It diagnoses the cause against your real schema and returns corrected SQL.

Make it faster

Ask which indexes your table is missing

It reads the real indexes and column stats, then writes the exact CREATE INDEX statements, each with the why and the trade-offs.

Know your data

Every table at a glance

Storage and index sizes, keys, indexes, engine and collation, plus the queries that recently touched the table.

Connect securely

Your databases, your keychain

Relational, document and vector backends, with optional SSH tunneling and AWS Secrets Manager. Credentials are encrypted by your OS keychain.

Bring your own key

Use your own AI provider, your key never leaves your machine

On Plus and Pro, use your own OpenAI, OpenRouter or Anthropic key. Requests go straight to that provider, so your key and prompts never touch our servers.