Backlog

Next up

In Progress

Done

Main Roadmap

Connect QueryFlow to your existing Informatica, AWS Glue, Airflow, or Talend instance. Claude reads every job definition and rebuilds it in QueryFlow's native format. The goal: cancel your enterprise ETL contract in an afternoon.

In Review

ETL migration from legacy tools — the kill shot

An Anonymous User

Optional iCloud-backed sync of your QueryFlow workspace across multiple Macs. Pipelines, saved queries, and schedules sync automatically. Credentials stay in each Mac's Keychain — never synced.

iCloud sync for saved queries and pipelines

Connect to databases behind firewalls via SSH tunnel — the standard pattern for production databases requiring jump host access. Supports password and key-based SSH authentication, multi-hop tunnels.

SSH tunneling for Postgres and MySQL

Currently Flow Books ship with pandas, numpy, matplotlib, requests, boto3 pre-installed. Custom pip install via Pyodide would let users add their own dependencies per workspace (e.g., openai, anthropic SDKs, custom ML packages).

Custom Python package install in Flow Books

Optional sharing of queries and pipelines with team members. Shareable links generate a read-only or editable view. Comments inline on specific cells. Designed for small teams where iCloud isn't enough but full enterprise collab is overkill.

Collaboration via shared queries and pipelines

After running a query, Claude AI analyzes the result columns and suggests the best chart type. One click to generate the matplotlib/plotly code in a new Python cell. Smart defaults based on data types and column count.

AI chart suggestions in Flow Books

Stream change events from Postgres logical replication or MySQL binlog instead of running scheduled snapshots. Sub-second replication SLAs for use cases that need them. Trade-off: meaningfully more complex setup than batch snapshots.

Real-time CDC mode for Postgres and MySQL

Automatic versioning of saved queries with diff view. See what changed between versions. Restore any previous version with one click. Lightweight SwiftData snapshots — no Git complexity.

Query version history with lightweight snapshots

Publish your Flow Book results directly to a Notion page or Google Sheets workbook. Cells render as proper Notion blocks (code blocks, tables, charts as images). Useful for sharing analysis with non-QueryFlow team members.

Export Flow Books to Notion and Google Sheets

Native Google BigQuery support alongside Snowflake and Redshift. Service account authentication, standard SQL dialect, scheduled queries with cost-aware billing tier selection.

Planned

BigQuery connector

Push query results or job status updates to Slack channels via incoming webhooks. Formatted tables using Slack Block Kit. Conditional alerts based on result thresholds (e.g., post only when at-risk deals exceed N).

Slack webhook destination

Read Contacts, Companies, Deals, Tickets from HubSpot. Write back records for marketing list imports and reverse ETL. OAuth 2.0 with full API scope.

HubSpot connector (source and destination)

Currently Google Sheets is source-only. Extending the destination model so pipelines can write results back to a Google Sheets tab. Includes append mode and replace mode.

Google Sheets as a write destination

Schedule jobs to fire relative to calendar events. Examples: '48 hours before every home game' for sports analytics, '1 hour after every standup' for team operations. iCal import plus live iCloud Calendar sync with event-matching rules.

Calendar-based job triggers with iCal/iCloud sync

SQLite support via GRDB.swift. Bundled directly into QueryFlow with no separate driver install. Useful for local development, sample data, and quick file-based projects.

SQLite as a bundled connector

MongoDB support via mongodb-nio or REST API. Document-aware schema explorer that handles MongoDB's flexible schema. Aggregation pipeline support in the editor.

MongoDB connector

Generic HTTP POST destination for any service that accepts JSON webhooks. Configure URL, headers, and payload format. Works with Microsoft Teams, Discord, Zapier, and any custom endpoint.

Webhook destination for custom HTTP POST

Extending Salesforce write operations beyond the Visual ETL field mapper — direct destination support in pipelines so any source can push to Salesforce without the mapper UX.

Salesforce as a write destination beyond field mapper

ClickHouse via the HTTP interface. Fast columnar query results with ClickHouse-specific syntax highlighting and EXPLAIN support.

ClickHouse connector

Native Databricks support via the SQL Statement Execution API. OAuth and PAT authentication. Works with both Unity Catalog and legacy workspaces. Lakehouse-aware schema explorer.

Databricks connector

Currently CSV/TSV files work as sources. Extending the SpreadsheetStrategy to also handle Excel .xlsx workbooks — read any tab, handle multiple sheets, type inference for date/number columns.

Excel .xlsx as a data source

Currently you can create schedules from SQL or Flow Books but the edit screen doesn't differentiate between source types. Adding context-aware UX so editing a SQL-sourced job looks different from editing a Flow Book or pipeline source.

Scheduler edit screen with context awareness

Surface Google Sheets-specific connection details (which Google account, which spreadsheets are accessible, OAuth scope status) in the connection detail panel. Same UX pattern as Snowflake and Salesforce connection details.

Connection detail view for Google Sheets

Push data into Salesforce from any source (Snowflake, Redshift, Postgres, MySQL, CSV). OAuth 2.0 authentication. Insert, Update, Upsert sync modes. Maps source columns to Salesforce field names with AI assistance.

Completed

Salesforce reverse ETL with OAuth and field mapper

Claude AI suggests SQL completions as you type, using full awareness of your connected database's schema. Tab to accept the suggestion. Works against any connected database — Snowflake, Redshift, Postgres, MySQL, Google Sheets, Salesforce.

Claude AI ghost text completions with full schema awareness

Mixed SQL and Python cells in one notebook. Query results auto-inject as a pandas DataFrame named df. Pyodide WASM runtime ships pandas, numpy, matplotlib pre-installed. Markdown cells for documentation.

Flow Books — SQL + Python notebooks with shared state

Drag-and-drop pipeline builder with curved connection lines between source and destination fields. Auto-map algorithm matches columns by name similarity. Claude AI 'Suggest Mappings' button for harder cases. Transform chips for UPPERCASE, lowercase, TRIM, date format conversion, custom Python.

Visual ETL pipeline with bezier-curve field mapper

Cron, interval, daily, weekly, and manual triggers. Jobs run via SMAppService (App Store compliant login item). Missed jobs catch up automatically on next wake. Observatory dashboard tracks every run with full history and logs.

Scheduled jobs that run while your Mac sleeps

Claude's chat panel now sees your last 5 query errors automatically, including timestamps and SQL snippets. When a query fails, ask Claude what went wrong and it analyzes the actual error message in context of your schema. The red ErrorContextChip surfaces in the chat when recent errors are present.

Claude AI is now error-aware

Connect to Google Sheets via OAuth 2.0 through your system browser. Pulls live data on every query — your Sheets edits show up immediately in QueryFlow. Sheets API v4 with token refresh handled automatically.

Google Sheets connector with live OAuth

Native Postgres connector using postgres-nio TCP wire protocol. Works with localhost, RDS, Aurora, Neon, Supabase, Heroku Postgres, and any standard Postgres deployment. Full SSL support.

PostgreSQL connector via postgres-nio

Native MySQL connector via mysql-nio TCP wire protocol. Supports localhost, RDS, Aurora, PlanetScale, MariaDB, and public no-auth servers (great for trying QueryFlow against the public Rfam MySQL database).

MySQL connector with public/no-auth support

Instead of HTTP 422 cryptic errors, QueryFlow now parses Snowflake's JSON error responses and surfaces the real error message, code, SQL state, and statement handle. Errors are selectable, copyable with one click, and include retry buttons.