PromptOT changelog · v1.17.0

PromptOT changelog,
release by release.

A complete record of updates across prompt management, versioning, evaluations, API delivery, and MCP tooling.

24
Releases
118+
Improvements
2
Years shipping
1.17.0
Current
Latest releaseSeptember 2026

Prompt workspace refresh

  • The prompt name, centered search, and Share and Save actions share the top header, with tags available beside the prompt name
  • Dedicated Blocks, Preview, and Compare views give the editor more room
  • Editor views, Playground model and Run controls, and test and evaluation actions share the tab row
  • Version status and History stay available across Editor, Playground, Tests, and Evals
  • The sidebar groups workspace navigation, theme, and account controls, with its collapse button beside the workspace breadcrumbs
  • Copilot keeps your conversation and unfinished input as you switch tabs or resize the workspace
  • Failed project requests now offer a retry instead of leaving the page loading
  • Playground controls adapt to available space, and dialogs and sidebar navigation remain usable on small and landscape screens
202622 releases
September 2026

Prompt workspace refresh

  • The prompt name, centered search, and Share and Save actions share the top header, with tags available beside the prompt name
  • Dedicated Blocks, Preview, and Compare views give the editor more room
  • Editor views, Playground model and Run controls, and test and evaluation actions share the tab row
  • Version status and History stay available across Editor, Playground, Tests, and Evals
  • The sidebar groups workspace navigation, theme, and account controls, with its collapse button beside the workspace breadcrumbs
  • Copilot keeps your conversation and unfinished input as you switch tabs or resize the workspace
  • Failed project requests now offer a retry instead of leaving the page loading
  • Playground controls adapt to available space, and dialogs and sidebar navigation remain usable on small and landscape screens
September 2026

Stronger sign-up protection

  • Sign-in now runs an invisible Cloudflare Turnstile check — most people never see it, and it only asks for a click when a request looks automated
  • New accounts confirm their email with the 6-digit code before the account is activated
  • The Prompt Library newsletter now uses double opt-in: we email you a confirmation link, and you're subscribed only after you confirm
  • The free Prompt Structurer and Prompt Quality Analyzer now run the same invisible bot check
September 2026

A more helpful welcome

  • Getting-started emails adapt as you create prompts, connect MCP clients, try co-pilot, and invite your team
  • Email greetings follow your updated profile name, and changing your email stops lifecycle mail to the old address
  • Returning to PromptOT ends the remaining check-in emails
  • Email links return you to the intended page after signing in or completing onboarding
  • Manage lifecycle email preferences with one-click unsubscribe
July 2026

Book Demo Call from the hero

  • Landing hero now offers a Book Demo Call CTA next to Get started free — opens Cal.com's native responsive scheduling popup
  • Real availability, timezone detection, .ics calendar invites, and reschedule/cancel links come from Cal.com — no hardcoded time slots or custom booking storage
  • Target event is configurable via NEXT_PUBLIC_CAL_LINK so sales can point the modal at any Cal.com event without a code change
  • Analytics events wired to Cal's iframe bus: book_demo_click, book_demo_opened, book_demo_ready, book_demo_confirmed
  • Cal.com owns the popup sizing, focus, scrolling, and close behavior, avoiding nested-modal layout conflicts across desktop and mobile
July 2026

Token Counter Refresh

  • Token Counter now compares 12 current models — GPT-5.6 Sol/Terra/Luna, Claude Fable 5, Opus 4.8, Sonnet 5, Haiku 4.5, Gemini 3.1 Pro, 3.5 Flash, 3.1 Flash-Lite, Grok 4.5, and DeepSeek V4 Pro — with July 2026 pricing verified against official vendor pages
  • New estimated-response selector: pick a short, medium, or long-form reply and per-call costs split into an input + output breakdown per model
  • Redesigned layout: a summary sidebar shows your prompt's token range across tokenizers, cheapest and priciest per-call cost, and the price spread between them at a glance
  • The cost table now ranks models cheapest-first with per-1k-calls cost bars, highlights the cheapest model, shows each model's context window and per-1M input/output rates, and collapses to the top six with a "show all" expander
  • Vendor long-context tiers (GPT-5.6 above 272K input tokens, Gemini 3.1 Pro above 200K, Grok 4.5 above 200K counting the estimated response per xAI's request-based billing) are applied automatically, prompts larger than a model's context window show as not fitting instead of a misleading price, and Claude Fable 5 / Opus 4.8 / Sonnet 5 counts include Anthropic's documented ~30% newer-tokenizer increase
  • Try-a-sample button, word count, and faster recounting (tokenizers are now cached and shared across models with the same encoding)
  • Related reading links to the blog and a sharper save-to-workspace call to action below the results
June 2026

Answer Engine Optimization (AEO)

  • New /llms.txt and /llms-full.txt endpoints so AI answer engines (ChatGPT, Claude, Perplexity, Gemini, Copilot) can discover and ingest the full evergreen catalog — glossary, use cases, comparisons — in a single fetch
  • Explicit robots.txt rules welcoming 16 AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, CCBot, and more) alongside the existing organic-search rules
  • Blog posts now ship BlogPosting + speakable structured data so voice assistants and AI search can lift article headlines and ledes directly
  • /docs/mcp/install gained HowTo schema with a five-step install flow and a totalTime hint — a high-intent target for queries like "how to install PromptOT MCP"
  • JSON-LD across 14 marketing pages is now rendered through a single shared component with universal </script> XSS escaping — previously inline and inconsistent
  • New aeo_referral analytics event fires when a visitor lands from an AI answer engine, with engine attribution across 12 sources (ChatGPT, Perplexity, Gemini, Copilot, Claude, You, Phind, Kagi, Brave, Meta AI, Groq, Bard)
May 2026

Per-Prompt Model Configuration

  • Choose the LLM model for each prompt right from the editor header — search the live OpenRouter catalog and pick any model
  • Tune temperature, max tokens, and top-p alongside the model, all delivered with the prompt via the API
  • Model settings are captured in version history: each saved version snapshots its model config, and the published version's model is what the delivery API returns
  • The version history panel now shows which model each version uses
April 2026

Free Developer Tools

  • New /tools hub with three free developer tools — no signup required
  • Token Counter: live token counts across GPT-4o, Claude Sonnet 4, Gemini 2.5 Pro and more, plus cost estimates per 1,000 calls. Runs entirely in your browser.
  • Prompt Structurer: paste an unstructured prompt and get it split into typed blocks — role, context, instructions, guardrails, output format
  • Prompt Quality Analyzer: instant 0–100 quality score with specific strengths, weaknesses, and actionable suggestions for any LLM prompt
  • Save any tool output to PromptOT via a one-click signup flow — the pasted prompt is hydrated into your first prompt automatically
  • AI-powered tools are abuse-protected with Cloudflare Turnstile, per-IP rate limits, and SHA-256 hash logging (raw prompts never stored)
April 2026

Pricing Page & SEO Polish

  • Dedicated /pricing page — Free, Pro, and Enterprise tiers with direct links from the navbar and footer (previously an on-page anchor)
  • Homepage now ships SoftwareApplication JSON-LD with tier offers so search engines can surface pricing in rich results
  • Keyword-targeted metadata and canonical URL on the homepage — previously it inherited generic defaults from the root layout
  • Twitter card site/creator handles added to default metadata so shared links attribute to @promptot
  • Breadcrumb schema on /pricing for richer navigation hints in search results
April 2026

MCP Server

  • PromptOT now works in Claude Desktop, Cursor, Codex CLI, claude.ai, ChatGPT, Windsurf, and Zed via the Model Context Protocol
  • 23 MCP tools across prompts, blocks, variables, versions, and test cases — list/create/update/delete/publish from inside your AI assistant
  • Generate a scoped MCP key from any project's API Keys page with one click — install snippets ready to copy for every supported client
  • Hosted OAuth flow at mcp.promptot.com for browser-based clients (claude.ai, ChatGPT) — sign in with PromptOT, pick your project, approve
  • AI features (playground, evaluations, AI rewrites, prompt imports, co-pilot) are opt-in via a checkbox so free-tier users don't accidentally burn AI credits
  • MCP-originated mutations are tagged in the audit log and request logs so you can trace exactly which tool ran what
April 2026

Playground Output Formats

  • Switch between Raw, Markdown, JSON, and HTML rendering for playground output
  • Markdown view renders headings, lists, tables, blockquotes, and code blocks inline
  • JSON view pretty-prints and syntax-highlights valid JSON, with a clear error state for non-JSON output
  • HTML view previews model-generated HTML inside a sandboxed iframe (scripts disabled) to safely inspect markup
  • Copy button continues to copy the raw output regardless of the active view
March 2026

Public Prompt Gallery & Sharing

  • Public prompt gallery at /library — browse community-shared prompt templates
  • Shareable prompt URLs at /library/{slug} with block-by-block preview and copy-to-clipboard
  • Embeddable prompt card API endpoint for documentation and READMEs
  • "Powered by PromptOT" attribution header on all API v1 responses
  • Tag-based filtering and search in the gallery
  • One-click toggle to make any published prompt public from the editor
March 2026

Reddit Pixel & Conversions API

  • Reddit Pixel integration for client-side conversion tracking across all pages
  • Server-side Reddit Conversions API for reliable attribution (sign up, prompt creation, prompt publishing, API key creation)
  • Multi-provider analytics — Reddit and Google Analytics now fire in parallel
March 2026

Audit Log

  • Full audit log page with filterable table of organization activity
  • Filter by action type, resource type, and date range
  • Paginated results with user details and event metadata
  • Color-coded action badges for quick visual scanning
February 2026

Editor Redesign & Error Tracking

  • Redesigned prompt editor with dual-scroll layout and fixed header
  • Card-based block layout with polished panels and visual cues
  • Sentry error tracking for frontend and API with source map uploads
  • SEO glossary and comparison pages for AI and prompt engineering terms
  • Auto-polling for running evaluation runs with live status updates
February 2026

Evaluations UI Improvements

  • Test case names displayed in evaluation results instead of truncated IDs
  • Clickable test case names open a detail modal with input, expected output, variables, and criteria
  • New evaluation result viewer modal showing actual LLM output, test input, and expected output side by side
  • Model stats (latency, token breakdown) shown in result detail view
  • Friendly model display names throughout the evaluations page
  • Fixed evaluation run creation failing when no specific test cases were selected
February 2026

Member Role Management & Co-pilot Test Case Generation

  • Generate test cases directly from the co-pilot — describe what to test and get domain-aware proposals
  • Balanced test mix: happy path, edge cases, boundary tests, and adversarial inputs
  • Per-test-case Accept/Reject with bulk Accept All / Reject All support
  • Variable-aware generation exercises different variable combinations from your prompt
  • Each test case includes 2-3 verifiable evaluation criteria (contains, regex, similarity)
  • Organization owners can now change member roles directly from the settings page
  • Inline role selector dropdown for quick role changes between owner, editor, and viewer
  • Automatic project access adjustment when promoting a member to owner
February 2026

Prompt Co-pilot

  • Conversational AI co-pilot for prompt editing — describe changes in natural language
  • Multi-block actions: edit, add, remove, reorder, and toggle blocks from chat
  • Inline diff previews with per-action Accept/Reject controls
  • Prompt quality analysis with scoring, strengths, weaknesses, and suggestions
  • Persistent conversation history per prompt with model selection
  • Prompt engineering best practices built into the co-pilot's knowledge
February 2026

Public Beta Launch

  • Launched PromptOT to the public
  • Landing page with full product overview
  • Documentation and API reference
February 2026

Access Control & Collaboration

  • Project-level access control for organization members
  • Role-based permissions (owner, admin, member, viewer)
  • Member invitation system
January 2026

AI-Powered Prompt Editing

  • AI block rewriting (improve, expand, simplify, custom modes)
  • Playground evaluation with OpenRouter integration
  • Prompt usage documentation with code snippets
January 2026

Webhook Notifications

  • Webhook delivery for prompt events (published, updated, deleted)
  • HMAC-SHA256 payload signatures
  • Exponential backoff retry (up to 5 attempts)
20252 releases
December 2025

API Delivery System

  • Public API for prompt delivery (/api/v1/*)
  • API key authentication with SHA-256 hashing
  • Environment scoping (production vs development)
November 2025

Block-Based Composition

  • Typed prompt blocks (role, context, instructions, guardrails, output_format, custom)
  • Drag-and-drop block reordering
  • Variable interpolation with {{key}} syntax
  • Prompt compilation engine

Want updates shipped to your stack?

Try PromptOT free. New features land here every week — no credit card required.

Get started freeSubscribe via RSS