Released, in the open.

AI providers

Two protocols, the models people use.

AI Chat and AI Translate now share one protocol-aware runtime. Choose OpenAI Compatible for GPT, DeepSeek, Qwen, Kimi, GLM, Gemini, and compatible gateways, or Anthropic Messages for Claude.

Claude, natively

Anthropic providers use the native /messages schema and headers, top-level system prompts, image content blocks, and streamed text, thinking, and token-usage events. Claude no longer depends on an OpenAI-shaped proxy.

Chat Completions and Responses, clarified

OpenAI Chat Completions and Responses are model-level endpoint choices inside the OpenAI-compatible family. This keeps provider setup simple while still supporting Responses-only models and OpenAI-native web and image tools.

Presets and safe migration

Provider presets cover OpenAI, DeepSeek, Qwen, Kimi, GLM, Gemini, and Anthropic. Existing configurations upgrade automatically: providers default to OpenAI Compatible, and the old Responses toggle migrates to the new model endpoint field.

Spotlight

Convert without opening the app.

DevUtilities now ships 11 Spotlight commands built on App Intents. Press ⌘Space, type what you need — convert a timestamp, generate a UUID, hash a string — fill the parameters inline, and read the result right in Spotlight. The app never has to open, and results are copied to your clipboard automatically.

11 commands, 9 tools

Convert Timestamp (smart two-way, type now for the current time), Convert Number Base, Convert Unit (all 7 categories, e.g. 1024 MB → GB), Encode/Decode Base64, URL Encode/Decode, Decode JWT, Hash Text (MD5/CRC32/SHA-1/256/384/512), Generate UUID (v4/v7), and Generate Random String.

Shortcuts and Siri included

Every command is a native App Intent, so the same actions appear in the Shortcuts app for automation and can be invoked through Siri. On macOS 26 they surface as inline Spotlight actions with quick-key support.

Tool naming

Shorter labels, clearer shelf.

Tool names are now concise and consistent across the app, website, and documentation. Recognizable technologies and formats use their canonical names — Base64, JWT, JSON, SQL, HTML, UUID, Crypto, URL, and Parquet — without repeating actions the tool already makes clear.

Clear where context matters

Ambiguous labels keep enough context: Unit Converter, Struct Converter, and Data Converter remain descriptive; HTTP Request is now HTTP Client, IP Query is now IP Lookup, and Base Converter is now Number Base.

No broken preferences or links

The change is presentational. Stable tool identifiers and existing website URLs remain unchanged, so saved sidebar settings and bookmarked pages continue to work.

Data Converter

One format in, another out.

The 25th tool converts between JSON, YAML, TOML, and CSV in any direction. Pick a source and a target format with a from/to picker, or hit the swap button to flip them. Conversion runs instantly and entirely on-device.

Order preserved, dates verbatim

A new shared value model stores objects as ordered key/value pairs and keeps date literals untouched, so conversions don't reshuffle your fields or reformat your timestamps. JSON is parsed with a custom order-preserving scanner, and YAML goes through the Yams library to preserve key order and quoting.

Nested data ↔ CSV

Nested objects and arrays flatten into dotted-key columns (address.city, tags.0) with an ordered union header across rows; reading CSV rebuilds the nesting. An Infer types toggle coerces cells like 123 and true into numbers and booleans, or keeps every cell as a string.

Shared pipeline

The Struct Converter now parses through the same value model — its output is unchanged.

Struct Converter

Sample data in, typed code out.

The 24th tool turns sample data into a ready-to-paste type definition. Drop in JSON, TOML, YAML, or a SQL CREATE TABLE block, and pick an output language — TypeScript, Python, Go, Java, Rust, Swift, or PHP. The tool walks every nested object, infers strings, numbers, booleans, ISO 8601 dates, arrays, and nullable fields, and emits idiomatic field names with serde renames, Swift CodingKeys, or Go struct tags preserving the original keys.

Four input formats

JSON parses via Foundation; TOML supports tables, dotted keys, and arrays of tables; YAML handles block-style mappings and sequences; SQL DDL parses multi-statement CREATE TABLE blocks and maps SQL column types (INT/VARCHAR/DECIMAL/TIMESTAMP/...) to language-native types, respecting NOT NULL.

Seven output languages

TypeScript interface, Python @dataclass, Go struct with JSON tags and aligned columns, Java POJO with getters/setters, Rust struct with #[derive(Serialize, Deserialize)], Swift Codable struct with CodingKeys, and PHP class with typed properties. Closes issue #17.

AI Chat enhancements

Markdown that renders the way you wrote it.

Enhanced Markdown rendering

Migrated to the Textual rendering engine for richer, more accurate Markdown output. Code blocks now have syntax highlighting across all languages, tables and nested lists render correctly, and streaming updates are smooth with no layout flicker.

One-click copy code snippets

Every code block in AI Chat responses now shows a Copy button. Grab any code snippet instantly without selecting text — perfect for copying commands, code examples, and configuration directly from the chat.

OpenAI Responses API support

Added support for the OpenAI Responses API, enabling compatibility with GPT-5-pro and future models. Includes full SSE streaming and reasoning/thinking output — see the model's chain of thought as it streams. Enable per-model via the Use Responses API toggle in model settings.

AI Translate — OpenAI TTS

Real-time text-to-speech, no waiting.

AI Translate now supports OpenAI's TTS engine alongside native macOS voices. PCM audio streams directly to the speaker — playback starts in realtime with no sentence splitting or waiting. Choose from 13 voices including alloy, nova, shimmer, echo, fable, and more. Configure mode (Auto / OpenAI / macOS) and voice in Settings → AI → Text-to-Speech.

Text Compare

Side-by-side comparison, with a real diff underneath.

A dedicated text comparison tool with visual diff highlighting. Compare any two text documents with real-time status updates, character and line metrics for both sides, and instant visual feedback showing additions, deletions, and modifications.

Diff mode for formatters

• JSON Diff. Visual diff editor for comparing two JSON documents side-by-side with syntax highlighting.
• HTML Diff. Compare HTML documents with automatic formatting and visual highlighting of differences.
• SQL Diff. Side-by-side SQL query comparison with automatic beautification.
• Unified Diff experience. All formatters share a consistent diff interface with real-time status updates.
• State persistence. All comparison views remember content between sessions.

Currency

Thirty-eight currencies, thirty days of history, no API hand-holding.

Real-time currency conversion across 38 major global fiat currencies — USD, EUR, GBP, JPY, CNY, KRW, INR, and 31 others. 24-hour smart caching keeps data fresh while minimising API calls; everything still works offline against the most recent cache.

• 30-day price history. Daily snapshots build a complete historical view automatically.
• 24-hour trend indicators. Visual up/down arrows with percentage change vs. yesterday.
• Flexible number input. Supports both formatted (1,000,000) and plain (1000000) numbers.
• Quick amount shortcuts. 1, 100, 1K, 10K — and a swap button for the pair.
• State persistence. Remembers your last conversion settings between sessions.