What's New in DBConvert Streams
Values that arrive intact. Runs that finish, or say why.
Version 2.6.0
Released August 1, 2026
A migration that reports success should be one. This release fixes conversions that changed values without saying so, and runs that stopped without saying why.
Values arrive intact
Money keeps its value
A PostgreSQL money column arrived as 123456 instead of $1,234.56. It now lands as DECIMAL(19,2).
CDC keeps the clock
Replicated timestamps were shifted by the host machine's UTC offset: a row written at 11:00 arrived at 13:00.
Wide tables keep their rows
A batch too large for the transport was dropped in silence — the run reported Finished over an empty target. Batch size is no longer a setting: batches are sized by what they hold.
Types stay themselves
A zero primary key is no longer renumbered, a NULL binary column no longer becomes empty, DATETIME keeps its precision, and long text no longer lands in an unmeasured VARCHAR(255).
Runs finish, or say why
No more silent stalls
A large load could stop making progress and stay Running for ever — and that one run then refused every later start until the service was restarted.
A failure says what happened
A failed run carries the database's own error instead of nothing at all, and an unreachable host now fails in a second, naming the connection.
Stopping is not failing
A run you stop reports Stopped and keeps its progress. Row-level failures no longer decide the run's outcome.
CDC checks the target first
An initial snapshot names the target tables that are not empty before the run, instead of failing mid-way.
Files, S3 and queries
Exports match the source
CSV, JSONL and Parquet, local or on S3: a column named order no longer fails the export, unsigned integers no longer wrap, decimals stay decimals and UUIDs stay UUIDs. Loading any of it back returns the values and the types it left with.
Your query runs on your database
A query against a single source now runs on that source — its SQL dialect, its column types, the result the editor previewed.
One slow query, one slow query
Queries across several databases moved out of the API into a service of their own, so two people querying at once no longer collide and a crash can no longer take the API with it. Large results arrive in pieces.
Upgrading
Because the types above are now mapped correctly, this version creates targets that differ from 2.5.x ones. Re-running an existing configuration leaves an existing target's schema alone — recreate the target to get the corrected one.
Docker deployments gain one container, stream-query-worker, which runs cross-database and file queries outside the API. Pull the updated compose file to get it; see Docker deployment.
Version 2.5.0
Released July 21, 2026
AI Chat
AI Chat is the assistant inside the DBConvert Streams desktop app. Ask about the database, file, stream, or SQL query in front of you. It reads the live workspace with DBConvert's read-only tools, shows what it inspected, and answers from real state.
Talk to your database
Ask what is in a database, which tables matter, or why a stream stopped. The assistant inspects the actual workspace before it answers.
Generate SQL
Press ⌘K / Ctrl+K in SQL Console, describe the query, then review the generated SQL with Keep or Revert.
Fix failed SQL
Use Fix with AI to attach the failed query and database error automatically, then review the replacement before keeping it.
Open AI Chat for a connection, ask a question, and see the tools it uses before it answers.
Each chat keeps the scope it was opened from. Choose an installed agent and model per chat; the header shows token usage and cost. Agents use only DBConvert's read-only tools — Chat does not grant shell, filesystem, or code-editing access.
Run monitoring
A config now owns its run history
Expand a stream config to see its runs in one place. Filter the history by Finished, Stopped, or Failed, then open the run that needs attention.
- The primary action follows state: Run Again, Resume Load, or Reset.
- Monitor shows progress, rows, chunks, rates, and per-table status for that run.
- Stopped loads keep their saved progress; a user stop is no longer reported as an error.
Version 2.4.2
Released June 12, 2026
CDC and workspace reliability
No more orphaned replication slots
Deleting a CDC stream config now also drops its PostgreSQL replication slot on the source. Previously the slot stayed behind and silently retained WAL — on a busy source that means unbounded disk growth long after the pipeline is gone.
The docs now also cover how slot WAL retention works and how to cap it with max_slot_wal_keep_size — see CDC reliability and resume.
Connection secrets stay in sync across services
Connections you add or update are now picked up immediately by every running service that shares the workspace — including the MCP server your AI assistant talks to. No restarts after adding a connection: create it in the app, query it from Claude or any MCP client right away.
Version 2.4.0
Released June 10, 2026
AI Assistants via MCP
DBConvert Streams 2.4 ships a built-in MCP server: connect Claude, Cursor, VS Code Copilot, Windsurf, Gemini CLI, or Codex, and your AI assistant reads live schemas, data, and stream state from the same workspace you use — no more pasting DDL and error logs into chat.
What you can do now
- Ask your assistant to describe tables, run read-only SQL, and join data across PostgreSQL, MySQL, and files in one federated query — 26 read-only tools in total.
- Diagnose streams in plain language: status, recent errors, logs, and throughput stats.
- Connect in one click from the new ✨ AI Assistants panel — 7 clients registered automatically, no account required on desktop.
- On Docker deployments, remote AI clients connect over HTTP(S) with your API key.
- Read-only by design: only
SELECTpasses the server-side filter — the AI can look and advise, never write.
The AI Assistants panel detects installed clients and registers the MCP server with one click.
Version 2.3.0
Released May 15, 2026
Paged SQL Console results
Large SELECT / WITH queries now return paged results, so an uncapped query can no longer overwhelm the grid or the backend.
- Page sizes: 50, 100, 250, 500, 1000 rows.
- Default 1,000-row cap for uncapped queries.
- More stable federated MySQL queries.
Version 2.2.0
Released May 11, 2026
Visual Query Plans in SQL Console
DBConvert Streams 2.2 adds visual Query Plans to the SQL Console, making it easier to understand how PostgreSQL, MySQL, and federated queries are executed.
What you can do now
- View query plans directly in the SQL Console.
- Compare estimated vs actual rows, timing, loops, joins, filters, sorting, and indexes where available.
- Switch between a clean visual Plan view and the original raw database output.
- Find performance signals faster with search and Signals only filtering.
- Use a new aggregate validation template for faster cross-database checks.
The Plan view highlights operators, indexes, row estimates, costs, and warning signals.
Select a plan node to inspect filters, joins, timing, loops, and other useful details.
Version 2.1.0
Released April 28, 2026
Safer Run Continuity for CDC and Load
Version 2.1 improves two separate ways to continue interrupted work:
- CDC streams with an initial snapshot now switch into live replication automatically after the snapshot completes, instead of requiring a manual handoff.
- Large supported Load runs can use Resume after Stop so the next start continues from the last saved point instead of restarting the whole load.
CDC Initial Snapshot Improvements
What changes for CDC streams
- Readiness checks: the stream verifies key conditions before the initial snapshot begins
- Automatic handoff to live replication: after the initial snapshot finishes, the stream continues into CDC without a manual operator step
- Easier restart before CDC begins: if the initial snapshot stops, starting the stream again is simpler and more reliable before live replication begins
The Replication Status stepper shows snapshot completion, handoff, and CDC apply while the stream is writing into PostgreSQL.
Resume After Stop for Large Loads
Large supported loads into MySQL or PostgreSQL targets can continue from the last saved point after a stop or interruption, so you do not have to restart the whole load from zero.
Resume from the saved point
When the run supports resume, the next start continues from the last saved point instead of repeating work that already finished.
When a load supports resume, the monitor keeps the saved state visible and lets the operator continue with Resume Load.
Reset when you need a full rerun
If you want a full rerun, Reset clears the saved state first so the next run starts cleanly from the source.
Editor and Connection Improvements
Improved SQL editor
Query results now open below the SQL code instead of in a side pane, and the results grid adds page-size controls and pagination for longer result sets.
The updated SQL editor keeps the query and its results in one vertical flow, with pagination controls at the bottom of the results grid.
Better Supabase connection guidance
Supabase setup is now easier to get right, with clearer connection guidance for direct access, session pooler use, and IPv6-related connection issues.
Version 2.0.0
March 2026
Two Ways to Run
Desktop Apps & Docker Web App
Version 2.0 ships in two distribution modes. Cross-platform desktop applications for Windows, macOS, and Linux — install and run locally with zero configuration. Or deploy the Docker-based web app on your own server for team access and production workflows.
Desktop App
- Native apps for Windows, macOS, and Linux
- One-click install — no Docker or server needed
- Full IDE features included free
- Local data stays on your machine
Docker Web App
- Deploy on any server or cloud provider
- Browser-based access for the whole team
- Seat-based licensing for shared environments
- Production-ready with Docker Compose
Rebuilt Data Explorer
A Completely New Explorer Experience
The minimal read-only viewer from 1.x has been replaced with a full-featured data exploration and editing environment. Browse, query, filter, and modify your data — all without leaving the UI.
The new Data Explorer: inline editing, SQL console, and multi-connection browsing in one view
Edit rows directly in the grid with visual change tracking — green for inserts, teal for edits, red for deletes
Inline Data Editing
Edit rows directly in the data grid. Insert new rows, update values, and delete records with full undo support and keyboard shortcuts. Changes are tracked visually with color-coded pending states before committing.
- Add, Edit, Delete: Full CRUD operations directly in the grid
- Visual Change Tracking: Color-coded rows for pending inserts, updates, and deletes
- Keyboard Shortcuts: Efficient editing with context menus and key bindings
- Undo Support: Revert changes before committing
Grouped Connection Filters
Filter connections by type (databases vs. storage) with match counts. Quickly narrow down large connection lists.
Universal Search
Search across connections, streams, databases, and file paths with highlighted matches and real-time result polling.
Reworked Navigation
Persistent section expansion, breadcrumb navigation, and a dual-pane layout for side-by-side database browsing.
Advanced Sorting & Filtering
Column sorting synchronized with query filters, custom filter panels, and AG Grid-powered data presentation.
Cross-Database SQL & Federated Queries
Multi-Source SQL Engine
Run SQL queries across MySQL, PostgreSQL, S3, CSV, and Parquet sources from a single interface. Mixed streams combine native table replication with ad-hoc analytical queries in one operation.
SQL console executing a cross-database JOIN between PostgreSQL and MySQL with live results
Unified Query Interface
- Cross-Source Queries: Join data across MySQL, PostgreSQL, S3, CSV, and Parquet in a single SQL statement
- Mixed Streams: Combine native table replication with analytical queries in one stream
- Federated Execution: Automatic routing of queries to the most efficient execution engine
SQL Console with Intelligence
Entirely new in 2.0 — a built-in SQL console for writing and executing queries directly in the UI. Includes real-time autocomplete, diagnostics, and signature help powered by an integrated Language Server that adapts to your connected databases.
Use SQL queries as stream sources — write a cross-database query, preview results, and stream the output
Console Features
- Built-in SQL Console: Write and execute queries without leaving the UI
- Autocomplete: Schema-aware suggestions for tables, columns, and functions
- Diagnostics: Real-time error detection and query validation
- Signature Help: Function parameter hints as you type
SQL Template Picker
Pre-built starter queries for common cross-database join patterns. Select a template, customize, and run — no manual SQL needed for typical workflows.
File & S3 Storage Support
Files and S3 as Sources & Targets
Version 2.0 introduces full support for file-based and S3-compatible storage as both sources and targets — a capability entirely absent in 1.x. Read from or write to CSV, Parquet, and JSONL files on local disk or S3, in both Convert and CDC modes.
Stream wizard with databases, S3, and local file connections available as both sources and targets
Storage Features
- Multiple Formats: Parquet, CSV, and JSONL supported as both source and target
- S3-Compatible: Works with AWS S3, MinIO, and other S3-compatible storage
- Convert & CDC: Full support for both bulk migration and real-time change capture to storage
- Mixed Streams: Combine native table operations with file/S3 targets in a single stream
Connection Wizard Redesign
Stream configuration with multiple sources, file targets, and format selection
Revamped Add Connection Wizard
Cleaner step-by-step flow with improved layout, enhanced parameter controls, and better visual feedback throughout the connection setup process.
CSV Dialect Detection
Automatic inference of delimiter, quoting, and encoding when connecting to CSV sources. No manual configuration required.
Stream Configuration & Control
Schema Policy & Write Mode
New controls for how target schemas are created (create, drop+create, skip) and how data is written (insert, upsert, replace).
Initial Load + CDC
CDC streams can copy existing rows into empty target tables before continuing from the matching MySQL binlog or PostgreSQL WAL position. MySQL now routes eligible bootstrap tables through resumable chunked copy automatically, PostgreSQL bootstrap remains plain, and the old manual resumable toggle is gone.
Simplified Pricing
March 2026
Free IDE + Seat-Based Plans
We've completely reworked our pricing. The IDE (Data Explorer, SQL Console, Diagram Viewer) is now free forever. No more volume-based tiers or data transfer limits — just a simple per-seat plan for streaming capabilities.
What changed
- No more data volume counting across plans
- IDE features are free — no plan required
- Simple per-seat pricing for Streams access
- Evaluation included: 500 MB Convert + 48h CDC
What's included free
- Data Explorer with inline editing
- SQL Console with autocomplete
- Database Diagram Viewer
- Connection management
Ready to try DBConvert Streams?
Install the desktop app and start with the database, stream, or question in front of you.