Tablething vs SQLly: wide Tauri client, deep SQL tool
I build SQLly, so I'm biased. Here's how Tablething, a $49 Tauri database client with 31 sources and BYOK AI, compares on engines, AI, dashboards and price.
Tablething showed up on Hacker News in June 2026 as a "local-first database client with BYOK AI". I build SQLly, so read this with that in mind. I'm biased, and I'll try to keep it honest anyway.
What I actually checked: I read the Tablething site, its pricing and FAQ, its GitHub README, its issue tracker and the v0.16.4 release files. I downloaded the Mac build and looked inside the app bundle without running it. I didn't use it against a real database and I didn't benchmark it. Its source isn't public, so anything about how it behaves comes from what it says about itself.
How they're built
Tablething is a Tauri 2 app. The back end is Rust, and the library names baked into the binary include sqlx, tiberius, DuckDB, the MongoDB driver and the Scylla driver. The interface runs in the webview your OS already ships. The Mac build is one universal binary, so it runs on Intel Macs too. It's signed, notarized, and updates itself. One developer makes it.
SQLly is also Rust, but it draws its whole interface on the GPU with no browser engine anywhere. That costs me download size in some places and wins it in others:
| Download | SQLly | Tablething 0.16.4 |
|---|---|---|
| macOS | 65.6 MB dmg (Apple Silicon only) | 76.4 MB universal dmg |
| Windows x64 | 60.9 MB zip | 22.9 MB setup .exe, 33.1 MB .msi |
| Linux x64 | 76.2 MB tarball | 41.9 MB .deb, 116.8 MB AppImage |
So on Windows and Linux it's a lot smaller. On the Mac it's a bit bigger, because it carries two architectures. And if you're on an Intel Mac, SQLly simply doesn't run there.
Databases
This is where Tablething goes wide. Its list has 31 sources, and it's not just SQL.
| Source | SQLly | Tablething |
|---|---|---|
| PostgreSQL, MySQL, MariaDB, SQL Server, SQLite | Yes | Yes |
| DuckDB, ClickHouse, Oracle, Turso | Yes | Yes (Oracle needs Instant Client installed) |
| Redis | Yes, with a full key and server toolset | Yes |
| CockroachDB, Supabase, Neon | Yes, as wire-compatibles | Yes |
| Snowflake, BigQuery | Planned | Yes |
| Redshift, Databricks, MotherDuck, Cloudflare D1 | No | Yes |
| MongoDB, Elasticsearch, OpenSearch, DynamoDB, Cassandra | No | Yes |
| CSV, Parquet, JSON files | Through DuckDB's read_csv and friends | Yes, each as its own source |
| Excel files | No | Yes |
If half your day is in MongoDB or Elasticsearch, that table already answers the question. SQLly connects to 30 engines, all relational, analytical or Redis, and I'd rather go deep on those than add a thin connector for everything.
The areas that matter
AI
Both apps keep AI on your terms. Tablething takes your own key for OpenAI, Anthropic, Google Gemini or Groq, or talks to Ollama or any OpenAI-compatible endpoint. It writes queries from plain questions, explains and fixes SQL, and has an agentic chat. Requests go straight to the provider you pick.
SQLly has no default provider at all. You run a local model through Ollama or MLX, point it at your own endpoint, or opt in to a cloud one. Agents in SQLly only read, and any write comes back to you as a proposal.
MCP
Both ship an MCP server. Tablething says its server gives Claude, Cursor and other clients read-only access without handing over credentials. SQLly's sqlly-cli mcp lets an agent read schemas and run a single SELECT inside a read-only transaction that always gets rolled back. Same idea, and I couldn't check how Tablething enforces its side since the code isn't public.
Dashboards and charts
Tablething lets you pin saved queries as dashboard tiles that refresh on their own. SQLly doesn't have dashboards. It charts and pivots a single result, and the "Server Dashboard" in SQLly is server health, not your data. If you want a wall of live tiles, that's a real point for Tablething.
Writing SQL and not breaking prod
This is where I spend most of my time on SQLly. Completion pulls real column values into a WHERE, writes the ON clause of a join from actual foreign keys, and a 0.5 MB script still edits at about 5 ms a keystroke. Results stream to disk, so memory stays flat at 50 million rows. Read-only connections, a hard stop on UPDATE or DELETE with no WHERE, and prod painted red come down from your environment settings. Then there's schema compare, data transfer, an activity monitor and backup/restore.
Tablething lists autocomplete, syntax highlighting, history, a spreadsheet-style grid, SSH tunnels and themes. I couldn't confirm a plan viewer, schema compare, admin tools, file export formats or read-only connections in the editor. That doesn't mean they're missing, just that its own pages don't say. To be fair to it, SQLly's plan viewer is still in progress too.
Cost
Tablething: a 7-day trial with everything unlocked, then $49 one-time for one device or $99 for three, with lifetime updates and an offline license key. Heads up, its README badge still says "free" and one of its comparison pages mentions free beta pricing, but the homepage pricing section is the current one.
SQLly: free for hobbyists with two saved servers, then $100 one-time for SQLly 1.x or $200 Lifetime. So Tablething is cheaper to buy. SQLly has a free tier you can keep using.
Neither is open source. Tablething's GitHub repo holds a README and release files, and its maker said on HN the plan is to open the core once it's stable. SQLly is closed too, with a source-escrow promise on Lifetime licenses.
Which one to pick
Pick Tablething if:
- You work with MongoDB, Elasticsearch, DynamoDB, Cassandra, Snowflake or BigQuery next to your SQL databases.
- You want dashboards of live query tiles.
- You query Excel, CSV or Parquet files a lot and want them as plain sources.
- You're on an Intel Mac.
- You'd rather pay $49 once than $100.
- You want a smaller install on Windows or Linux.
Pick SQLly if:
- You live in SQL Server, Postgres, MySQL or SQLite and write a lot of SQL by hand.
- You want guardrails that stop a bad
DELETEbefore it runs. - You need schema compare, data transfer, activity monitoring or backup scripts.
- You pull huge result sets and don't want the app to eat your RAM.
- You want AI to run on a local model by default, or not at all.
- You use Redis seriously, not just as one more source in a list.
One more thing worth knowing: Tablething is young. Its latest release, v0.16.4, is from 12 September 2026, and GitHub only keeps that one release. Have a look at its site for recent news before you commit.
The full side-by-side table is at /compare/tablething. If you want to try SQLly against the same database, it's at /download, free, no account.