Database tools

You can

• Filter with Search keys…. When you stop typing for 350 ms the app sends SCAN … MATCH *text*, a substring match done by the server.
• Read each key: an icon and a badge for its type (string, list, set, sorted set, hash, stream, or otro for a module type such as RedisJSON), TTL: (the seconds left, No expiry or Key no longer exists; red when 1 to 60 seconds are left) and its memory size, or size not reported.
• Tap a key to open a preview of its value, cut at 200 characters (up to 40 lines on screen). By type: a string reads GETRANGE 0..199, a list its first 10 items, a set 10 random members, a sorted set 10 items with scores, a hash the first 10 fields (HSCAN) and a stream 3 entries.
• Change a TTL: the timer icon (Set TTL), seconds in TTL (seconds), then Save. A negative number removes the expiry (PERSIST).
• Delete a key: the trash icon (Delete). It goes through the Dangerous Operation dialog showing DEL <key>; Execute Anyway confirms. The app tells you whether the server removed the key or it was already gone.
• Tap Load more to get the next page. A page is 200 keys, and a page can take up to 40 SCAN rounds of 250 before it stops.
• Open the Redis CLI (terminal icon) or the Redis Monitor (monitor icon).

You can't, and why

• Change a value, create a key or rename one from the screen: there is no control for it. Use the CLI (SET, RENAME…).Not yet
• Choose the logical database (SELECT n) in the browser: it comes from the database number saved in the connection.By design
• See a key's size when the server doesn't support MEMORY USAGE or the ACL forbids it: the card says size not reported instead of drawing 0 B.Third party
• Read a whole value in the browser: the preview stops at 200 characters. Use the CLI for the rest.By design

You can

• Type a command in the field (PING, GET key, SET key value…) and run it with the play button (Run).
• Use the four shortcut chips: PING, KEYS *, INFO and DBSIZE. KEYS * asks for confirmation like any command that can block the server.
• Confirm dangerous commands in the same Dangerous Operation dialog, with the reason in Spanish (the app writes it in Spanish in every language). They are: FLUSHALL, FLUSHDB and SWAPDB; deletes and overwrites (DEL, UNLINK, RENAME, GETDEL, MOVE, COPY, LREM, SREM, HDEL, ZREM, SPOP, LPOP, RPOP, XDEL, XTRIM…, the *STORE commands and SORT … STORE); server administration (CONFIG, ACL, CLIENT, DEBUG, SHUTDOWN, REPLICAOF, CLUSTER, SCRIPT, FUNCTION, MODULE, RESET…, except read-only subcommands such as CONFIG GET or ACL WHOAMI); commands that block the server (KEYS, MONITOR, BLPOP…); and EXPIRE, PEXPIRE and EXPIREAT.

You can't, and why

• Save favourite commands or scripts: there is no control for it.Not yet
• Subscribe to channels from the CLI: a subscribed connection can't run normal commands. The monitor opens a connection of its own for that.By design

You can

• Read Memory, Clients, Ops (operations per second, shown as /s) and Version.
• Watch Events: each published message with its channel, its text (cut at 500 characters) and the time. The list keeps the last 200.
• Stop it with Stop and start it again with Start Monitor.

You can't, and why

• See the commands other clients run: the monitor doesn't use Redis's MONITOR command, only INFO and published messages.By design
• See how the numbers changed over time: only the current values are shown.Not yet

You can

• Choose a project when your Google account has several: Choose a Firebase project. The header shows Project: … and Change project.
• Type a path in Collection path (for example users) and tap the arrow (Open path). A collection has an odd number of segments (users, users/alice/posts) and a document an even number (users/alice).
• Tap a document to open its subcollections (Google route). The back gesture goes up one level before it leaves the screen.
• Read documents: up to 50 per collection, each card with the document id, its first 4 fields and +N more fields.
• On your own project, edit a document with the pencil (Edit): Edit Document opens a JSON field and Save writes it. Delete one with the trash icon; it goes through the Dangerous Operation dialog showing DELETE <path>.
• Tap the filter icon (Firestore Query) to build a query.

You can't, and why

• List the root collections, or a document's subcollections, on the SDK route: Firestore's client SDK has no call for it. Type the path of a collection instead.Third party
• Edit or delete on the Google route: it is read-only, so the edit and delete buttons aren't drawn. Set the project up in Integrations to write.By design
• Create a new document from the browser: it only edits and deletes existing ones.Not yet
• See more than 50 documents of a collection in the browser: use Firestore Query, which goes up to 500.By design
• Read a named Firestore database: only (default) is read, on both routes.By design

You can

• Type the Collection path. If you leave it empty, the path you were browsing is used.
• Tap Add Clause under WHERE Clauses. Each clause has a Field, an operator chip (==, !=, <, <=, >, >=, array-contains or in) and a Value. An icon removes a clause.
• Values are typed for you: null, true, false, integers, decimals and ISO dates (2026-10-04 or 2026-10-04T10:00:00Z) are recognised, and anything else is text. Put a value in quotes ('123') to force text. With in, separate the values with commas and quote a value that contains one.
• Set Order by (a field) and Limit (25 by default; the app keeps it between 1 and 500).
• Tap Run Query. Under documents you get the count and one card per result with all its fields.

You can't, and why

• Sort descending: the screen has no switch for it, so results always sort ascending.Not yet
• Query a collection group (every collection with a given name): there is no control for it.Not yet
• Combine inequalities on different fields, or sort by another field, without an index: Firestore requires a composite index, and its error, with the link to create it, is shown as it is.Third party

You can

• Connection: Latency (the milliseconds of one SELECT 1 on your connection, measured when you open or refresh the monitor), Connections (the open connections to the database, on the local PostgreSQL only), Queries per second (always —), Queries Today and Avg Query Time (from the queries you ran in this app since midnight).
• Server sessions, on the local PostgreSQL only: each client session of pg_stat_activity (the monitor's own is left out) with its pid, state, wait event, Database, Application and its query (the first 200 characters). Two icons act on a session: Cancel query (pg_cancel_backend: it stops the query and keeps the session) and Terminate session (pg_terminate_backend: it closes the connection).
• Queries from this app: the queries of the current session, each with its state and duration. The icon Kill Query cancels one that is running and Clear Completed empties the finished ones.
• Opened from the start screen, with no connection, the sessions come from the local engine.

You can't, and why

• See the sessions of Neon, Supabase, MySQL, Redis, Firestore or MongoDB: the module reads them only from the embedded PostgreSQL. Anywhere else it says “… does not report open sessions — only the embedded engine does.”By design
• Get queries per second: it isn't measured, so the card shows —.Not yet

You can

• Choose the Slow threshold with the chips 100 ms, 500 ms, 1000 ms and 5000 ms (500 by default). The list is rebuilt when you open the monitor or tap the refresh icon, so refresh after you change the threshold.
• Each entry shows its duration (red above 2000 ms), the time, the SQL (up to 200 characters, 3 lines) and the table.
• With nothing to show it says No slow queries detected.

You can't, and why

• Keep the threshold after the app restarts: it isn't saved.Not yet
• See slow statements that did not come from this app: the log is made of your own audit entries, not of the server's.By design

You can

• Read each index: its name, its table and columns, a UNIQUE badge when it is unique and N uses (est.).
• N uses (est.) is an estimate, not a counter: it counts the entries of your audit log for that table whose SQL mentions the index's first column. The badge is green above 0.
• With none it says No indexes.

You can't, and why

• See indexes on MySQL, SQLite or other engines: the tab stays empty.Not yet

You can

• Tap Add Alert, give it an Alert Name, choose an Alert Type and a Threshold. Switch a rule on or off, or delete it, with its controls. N rules shows how many there are.
• An alert fires when the current value is above the threshold. Fired alerts appear in the red strip at the top of the monitor.

You can't, and why

• Get a notification with the app closed, or while you are not looking at the monitor: nothing runs in the background. Rules are checked only when you open or refresh it.Not yet
• Alert on Database Size where the engine doesn't report a size: that rule is skipped, not fired with a zero.Third party

You can

• Read each finding: its level, its title, the entity it Affects, the remediation and a Learn more link.
• On a row level security finding, tap Open in Policies to jump to Security → Policies.

You can't, and why

• Fix a finding from here: the tab only reads Supabase's advisors.By design

You can

• Read each query with its table, its Time, its Count, the Rows read and the Rows examined, as the API gives them (the time is in microseconds).

You can't, and why

• See PlanetScale's figures in milliseconds or merged with the Slow Log: the two sources use different units, so they stay separate.By design

You can

• Switch dashboards with the tabs. For each one you can Rename it or Delete it; deleting asks “Delete “X” and its N widgets? This cannot be undone.”
• Read the widgets in 1 column on a phone in portrait, 2 columns elsewhere and 3 columns on screens 1,200 dp wide or more. On a one-column grid, a 2-column widget simply fills the row.
• Refresh one widget with its own button; each one shows when it was read (Read at …).
• A widget can also say: The connection this widget was created with is no longer available., This widget has no query., The query returned no rows., No numeric column to plot. or the engine's own error.

You can't, and why

• Move a dashboard to another device: they live only on this one.Not yet

You can

• In the sheet, set the Widget title, see the SQL that will be saved (shown there, not editable), pick the target Dashboard or New dashboard, and choose the Shape, the Width and the Auto-refresh.
• Width: 1 column or 2 columns. Auto-refresh: Off, or every 5, 15, 30, 60 or 300 seconds while the dashboard is on screen.
• Each widget has a title, a refresh button and a menu (Widget options) with Edit widget, Move up, Move down and Delete. Edit widget changes the title, the SQL, the shape, the width and the interval.

You can't, and why

• Run a statement that writes: a widget runs on its own and on a timer, so only SELECT, WITH, EXPLAIN, SHOW and DESCRIBE run. A writing statement stays unexecuted with “Not run: UPDATE writes, and a dashboard only runs read-only queries.”By design
• Use :name parameters in a widget: it has nowhere to get the value from when it refreshes by itself.By design
• Drag widgets to reorder them: use Move up and Move down.By design
• Refresh faster than every 5 seconds: a widget that re-queries every second is a SELECT loop on someone's database.By design
• Use a widget on Redis, MongoDB or Firestore: they have no SQL, so the widget shows the engine's error.By design

You can

• Read Sessions, Commits, Rollbacks, Cache hits, Data size, Version, Deadlocks and Checkpoints. They refresh every 5 seconds while the tab is on screen.

You can't, and why

• See this tab on any other engine: it comes from the embedded PostgreSQL, which runs only on 64-bit ARM phones.Device
• Edit or save this tab: it holds no query.By design

You can

• Pick Bar, Line, Pie, Scatter or Heatmap (the chart type names are fixed English text).
• The X axis is the first column. The Y axis is the first numeric column after it, and the pie uses the same two columns with only positive values. Scatter crosses the first two numeric columns (with just one there is nothing to draw). Cells that aren't numbers are drawn as 0.
• Switch the legend and the grid with the two icons next to the chart (Toggle Legend and Toggle Grid).
• The refresh icon in the header redraws the chart from the console's last result.
• Tap the save icon (Export Chart) to write a PNG with a white background through the system file picker.
• Tap the chart icon (Add to dashboard) to keep the query as a widget.
• With no numeric column it says “No chart data. Execute a query with numeric columns first.”

You can't, and why

• Plot more than one measure at once: one numeric column goes on the axis, because several on a shared axis can't be read (hundreds next to units).By design

You can

• See the detected format as a badge, the count (N rows · M columns, only when the whole file was counted) and the first 50 rows under Preview Data.
• Edit Table Name. It starts as the file's name, cleaned up (clientes-2026.xlsx becomes clientes_2026).
• Tick or clear First row is the header (on by default; it only matters for CSV, TSV and XLSX), Create table (on by default: it runs CREATE TABLE IF NOT EXISTS) and Truncate first (off by default: it runs DELETE FROM the table first, and turns red when ticked).
• Read the Detected Types chips. They come from the 50 sample rows (“Types are deduced from the preview rows, not from the whole file.”); Create table deduces the types again from the whole file.
• Tap Import. A message tells you how many rows went into the table.

You can't, and why

• Import into a database that already has 4 tables on the Free plan, even into an existing table: the limit is checked before the file is read, and the Database Pro window opens.Pro
• Import .xls, .ods or any format outside the six, or read a sheet of an Excel file other than the first.By design
• Import into Redis, MongoDB or Firestore: they don't speak SQL.By design

You can

• Copy a dump, open the menu and tap Import from Clipboard. The text is split at each ; and the statements run one by one, with no transaction. If some fail, you see how many failed and the first three different messages.
• The import is written to the Audit log as IMPORT.

You can't, and why

• Keep a semicolon inside a string or a function body: the dump is split at every ;, so those statements break.By design
• Import a dump that creates tables on the Free plan once the database has 4: the Database Pro window opens first.Pro

You can

• Read each entry: the SQL (3 lines), the number of rows or an error icon, the duration, the connection's name · the date and the error.
• Tap an entry to load its SQL into the console.
• Tap Clear All to empty the list. It doesn't ask first.
• Queries with parameters are stored as the template. The derived credentials of Turso and Upstash are blanked out of the text before it is saved.

You can't, and why

• See a query's result: only the SQL and its figures are kept, whatever the tool's one-line description says (“Everything you have run, with its result”).By design
• Search, filter or delete a single entry: there is only Clear All.Not yet

You can

• Tap a query to load it into the console. The bin icon (Delete) removes it with no confirmation.
• Keep queries in folders: New folder (Folder name), the chips All, No folder and one per folder, and Move to folder on a query (it appears once you have a folder). Delete folder removes only the folder: “Deleting the folder keeps the queries: they move out of it.”
• On a Supabase project, a From Supabase section lists the snippets of your Supabase dashboard, read-only, with Open in console and Save a copy.

You can't, and why

• Save a 6th query on Free: the limit is 5 saved queries (Save a copy counts too).Pro
• Edit a Supabase snippet here: they are read-only. Open one in the console or save a copy.By design

You can

• Tap Add Masking Rule, type a Column pattern and choose a Mask type. Switch each rule on or off, or delete it.
• The pattern is a piece of text that must appear in the column name, ignoring case: mail masks email and mail_backup. It is not a regular expression, so email|password would look for that exact text. The first enabled rule that matches wins.
• A masked column shows the masked value in the data viewer. To see the real value, open the Columns sheet and use Show in clear (Show in clear (fingerprint) when the biometric lock is on). It lasts until you leave the table.

You can't, and why

• Mask the SQL console, exports, copies or what the AI chat reads: masking applies only to the data viewer's grid.By design
• Use a regular expression in the pattern.By design

You can

• PostgreSQL (the local engine, Neon, Supabase or a server): each role of pg_roles, without the system pg_* ones, with its privileges and the tables of the public schema they reach.
• MySQL, MariaDB and PlanetScale: the accounts, from mysql.user or, when that isn't readable (PlanetScale hides it), from the table privileges, with their tables.
• MongoDB Atlas: the database users of the project, read-only, when a cluster is selected.
• Any other engine says “This engine has no users or roles.”

You can't, and why

• Create, edit or delete roles here: the tab only reads.By design
• See an Atlas user's password: Atlas doesn't expose it.Third party

You can

• See each table marked RLS on or Unprotected and, for each policy, its name, command, roles, USING and WITH CHECK.
• A table with RLS off says “Row Level Security is off: anyone with this connection can read every row.” With RLS on and no policy it says only roles that bypass RLS (the table owner, service_role) can read or write it.
• Tap Copy SQL to copy an ALTER TABLE … ENABLE ROW LEVEL SECURITY or a CREATE POLICY … that you run yourself.
• You also get here with Open in Policies from Supabase's Schema health, or with an Open in Database action from the Backend Manager.

You can't, and why

• Run anything from this tab: “Nothing here executes: Copy SQL only copies.”By design
• See policies on MySQL, SQLite or the NoSQL engines: Row Level Security exists only in PostgreSQL.Third party

You can

• Tap Grant (Grant Permission), type the Role, pick the Table and tick the privileges under Select privileges: SELECT, INSERT, UPDATE, DELETE, CREATE, DROP, ALTER, INDEX or ALL PRIVILEGES. It runs GRANT … ON … TO ….
• The Revoke icon of an entry in that list runs REVOKE for the privileges the entry holds. Role and table names must be plain identifiers (letters, digits and underscores, not starting with a digit).

You can't, and why

• See or revoke the permissions that already exist: the list shows only what you granted in this session, so Revoke reaches only those. Use Roles to see the real ones.Not yet

You can

• With it on, opening the module shows Database locked (“Authenticate to open your databases”) and asks for your fingerprint or face, or the device's PIN, pattern or password.
• With it on, Show in clear (fingerprint) asks again before it reveals a masked value.
• With the lock off, the list says Free on every line and the note “With the lock off, none of these ask for anything.”

You can't, and why

• Count on the lock to confirm deleting a table, revealing a saved password or opening a PRODUCTION connection: the screen lists them as Asks, but the app doesn't ask yet.Not yet

You can

• Search with Search projects… (it appears only when there is more than one project).
• Read each project: its name, its status · region, its id in monospace and a green dot when it is active.
• Tap one. Neon, PlanetScale, Turso and Supabase open the project sheet; D1 goes straight to its tables; Atlas opens the connection form (see MongoDB); Upstash opens as a Redis database when it returns a password, and otherwise through REST as a read-only table.

You can't, and why

• Create or delete projects: the list only shows them. With none it says “Create one from the provider dashboard”.By design