v0.18.0

Added

• MongoDB support: connect to a mongodb:// or mongodb+srv:// deployment and
  explore it in a dedicated document shell. A database → collection tree lists
  collections with estimated counts and view / time-series badges; selecting one
  streams its documents into a continuously scrolling grid (a column per
  top-level field) that fetches a window at a time and never loads a collection
  whole. An extended-JSON find filter narrows the view, and the inspector shows a
  document as pretty-printed extended JSON that preserves BSON types (ObjectId,
  dates, decimals, binary).
• MongoDB analysis panels: a Schema panel samples documents to show each field's
  path, type distribution ("string 82% / int 18%"), and how often it is present;
  an Indexes panel lists keys and unique / sparse / ttl / partial properties; a
  Query panel runs an aggregation pipeline into a results grid; and Explain
  reports the winning plan, the index used, and documents examined / returned.
• MongoDB editing (writable connections): edit a document with a field-by-field
  Form editor — editable names, a searchable type picker, collapsible nested
  objects and arrays, _id kept read-only — or a raw extended-JSON editor.
  Insert, delete, clone, and drop are available; destructive operations sit
  behind an explicit confirm and every write is refused on a read-only
  connection.
• MongoDB workspace: a tabbed workspace like the SQL and Redis browsers.
  Collections open in reorderable, pinnable tabs (the same collection can open
  several times, each with its own filter and edits), a split view shows two at
  once (⌘), and Table / List / JSON render modes switch how documents display.
  Standard window chrome — footer status bar, collapsible sidebar (⌘B), docked
  AI panel — plus full keyboard and vim (hjkl, g/G, Ctrl-d/Ctrl-u) navigation.
• MongoDB AI assistant (⌘L): grounded in the connection, it inspects the
  deployment, profiles schemas and type drift, samples and queries documents,
  runs aggregations, and explains queries to flag missing indexes; at the write
  tier it proposes document, index, or collection operations to approve behind
  the same gates as the manual path.
• Searchable, grouped History dock: the left History panel (SQL and Redis shells)
  gains a live search box and collapsible grouped sections — SQL by Today /
  Yesterday / Earlier, Redis by "Recently viewed keys" and "Commands" — each
  showing its row count and force-expanding on a match.
• Compare tables (data diff): a new "table: compare against…" command reports
  which rows are added, removed, or changed between two tables, aligned by the
  left table's primary key, as a full-screen read-only report (summary, filter,
  changed cells shown old → new). Both tables are read key-ordered and streamed,
  so neither loads whole; it never writes.
• red mcp <connection>: a headless stdio MCP server. Point Claude Code (or any
  MCP client) at red mcp my-connection for Red's read-only database tools
  (schema, describe, profile, SELECT, explain) grounded in that connection, with
  no GUI and no ports. Writes are withheld and a tool-call budget bounds a
  runaway client, like the in-app MCP path.
• Redis batch console: a Line / Batch toggle adds a multi-line composer that runs
  many commands at once with per-command output, a live "running N / M" readout,
  and a Stop button. A destructive command is confirmed once up front rather than
  per line, and each still passes the read-only and destructive gates. Load a
  .redis/.txt file or save the buffer back out.
• Connect through a proxy: a new "Connect via proxy" section (network engines)
  reaches a database via a SOCKS5 or HTTP CONNECT proxy, with optional auth whose
  password is stored in the OS keychain. A connection uses either a proxy or an
  SSH tunnel, not both.
• Import connections from more tools: alongside DBeaver and DBGate, the import
  wizard now reads JetBrains DataGrip / IntelliJ (dataSources.xml),
  RedisInsight, and plain credential files (~/.pgpass, ~/.my.cnf,
  ~/.pg_service.conf). Passwords are imported when recoverable and otherwise
  flagged for re-entry, never silently dropped.
• Vim navigation: an optional keymap setting adds hjkl, g/G, 0/$, and
  Ctrl-d/Ctrl-u motions to the result grid, schema tree, and history dock,
  alongside the arrow keys. Off by default; applies live.
• Remove all RED data: a "Remove all RED data" action (Settings → Behavior, the
  palette, or red reset) deletes RED's config and cached-data directories and
  every keychain secret (connection passwords, SSH keys, AI keys) in one step. It
  shows what will be removed, is irreversible, and leaves the binary untouched.

Changed

• Delete/destructive confirmations are unified across the SQL, Redis, and MongoDB
  shells under one setting. Deleting a Redis key or a MongoDB document now asks
  first like a destructive SQL statement, and every confirm dialog carries a
  "Don't ask again" checkbox (Settings → Query → "Confirm destructive
  operations", on by default, turns it back on). The MongoDB drop-collection
  confirm stays server-gated and always asks.