Synero

Changelog

What's new in Synero

New features, improvements, and fixes, shipped fast.

·v2.33improvement

A heads-up for image requests

  • Ask Synero to make or animate an image and the composer now says it answers in text, with a link to PixelDojo, before spending a run or a free question. Ask anyway sends it as usual
·v2.32feature

GPT-6.1 Terra and Gemini 4 Argon

  • New models: GPT-6.1 Terra, OpenAI's mid-size model at half the price of GPT-6.1 Sol, and Gemini 4 Argon, Google's newest top model. 22 models in all
  • Pick either one in the composer or by id through the API (gpt-6.1-terra, gemini-4-argon). The Solo, Duel, Council and Deep presets are unchanged for now
·v2.31improvement

Ad cookies are now your choice

  • A banner asks before Meta's ad cookie loads. Decline and nothing about your visit is sent to Meta, from your browser or from our servers
  • The Privacy Policy now says exactly what is sent to Meta after you accept: page views, sign-ups, your first free question (never its content) and new subscriptions, with your email and account ID only as SHA-256 hashes
·v2.30fix

Faster, sturdier chats and billing from a second code audit

  • Answers start sooner, the chat and public pages load less JavaScript, and attached photos are resized before upload, so a phone photo over 5MB now attaches
  • Stop, Regenerate and errors behave: a stopped or failed regenerate brings the previous answer back, an answer cut off mid-stream says it is incomplete and is left out of the consensus, and a question the server turns away goes back into the box with the reason
  • Stopping a question charges for what the models already read and wrote, never for a combined answer or agreement check that had not started. A stopped free trial question uses that question; one that fails before any answer gives it back
  • Chat history no longer skips chats after deleting some or a failed Show more. Reopened chats keep their agreement reports, web search sources and cut-off notes, and the address bar follows the open chat so a reload reopens it
  • A billing hiccup now asks you to try again instead of showing an out-of-credit paywall. Credit pack buyers see their balance in the header, subscribers out of credit can add $5 right in the chat, and the billing page shows ended and ending plans clearly
  • If the checkout will not load and you pay on whop.com instead, the credits now land on your Synero account even when your Whop email is different
  • Voice input is charged from the audio length (about $0.01 per minute) and shows as Voice input in your credit history, and the agreement check has its own Agreement check line
  • Safer accounts: sign-out clears the browser even when the auth service is down, sign-in links can no longer send you to another site, an expired reset link offers a new one, and signing up with an email that already has an account says so
  • Email links (confirm sign-up, sign-in link, password reset) now open a Continue page and sign you in when you press Continue. Work email security scanners that open links in advance no longer use them up, and a link someone else sends you can no longer sign you into their account
  • A template link from someone else shows its instructions before you apply them, and each account can keep up to 100 templates
  • Better on phones and with a keyboard or screen reader: 44px tap targets, no zooming in on form fields on iPhone, readable light mode text, a Skip to content link, arrow keys for mode presets, and Escape no longer stops a streaming answer when it closes a menu
  • The ad pixels no longer read button text or emails typed into forms, and the Privacy Policy now lists them along with the Crisp support chat
  • Shared chats show the agreement report even without a combined answer, list the AIs in chat order, and keep working through a database hiccup
  • Pay as you go credit packs from $5 with no subscription are now described on the homepage, pricing and comparison pages, and in the Terms and Refund Policy
  • API: clear 400 errors for oversized or malformed input (context up to 20 items, synthesisInstructions up to 10,000 characters, boolean web_search), council-meta carries the queryId, a follow-up always joins its parent's thread, and long format: "json" runs no longer time out at the CDN
  • MCP: results list the models that failed and why there was no consensus answer, and no longer suggest a repeat call that would bill again
·v2.29improvement

Follow-ups from the API and MCP group into threads

  • A follow-up asked through the API or an MCP client (parent_query_id) now lands in a thread with the question it follows, so the conversation reads as one thread in your history

Earlier updates

·v2.28fix
Credits hold up under parallel questions
  • Each question reserves its likely cost before it starts, so a burst of questions sent at once can't run your balance below zero. If questions already running cover the rest of your credits, the next one asks you to wait for them
  • A refund and a chargeback on the same purchase can no longer take back more credits than that purchase added
·v2.27fix
Smaller fixes from the code audit
  • Screen readers announce when each answer and the combined answer are ready, Esc stops an answer from anywhere on the page, and the account and mobile menus work from the keyboard
  • If your subscription ended with credits left, Synero says so and offers to resubscribe instead of sending you to update billing
  • The MCP server (2.1.1) reconnects cleanly after a failed call and no longer blames your API key for network errors
  • API: malformed MCP requests and unknown ids get clear 400 and 404 errors instead of 500s, and an account holds up to 25 active API keys
·v2.26fix
Fixes from a full code audit
  • A chat whose connection drops mid-answer now says so instead of loading forever, and slow Deep runs keep their connection open
  • Pasting a second screenshot works, dropping several files at once respects the attachment limits, and a rating only shows as saved once it is
  • Faster pages: math rendering loads only when an answer has math, the homepage loads less, and long answers stream more smoothly on phones
  • Usage and billing pages count every question, however many you have run, and never show a subscriber as unsubscribed after a hiccup
  • Payments run through Whop only
·v2.25feature
Claude Haiku 5.5 replaces Haiku 4.5
  • Claude Haiku 5.5, released today, is now Anthropic's fast model in Synero: adaptive thinking, a 1M token context window, and a tenth of Haiku 4.5's price ($0.10 / $0.50 per million tokens; prompts over 100,000 tokens bill at 5x)
  • The low Council tier and the Anthropic low synthesizer run Haiku 5.5. Saved picks, presets and API calls that name claude-haiku-4-5 run on Haiku 5.5 automatically
·v2.24feature
Synero suggests how many AIs a question needs
  • As you type, the composer suggests a mode for your question: Solo for simple ones, Duel when some judgment is involved, Council when experts would disagree, Deep when the stakes are high. One click applies it
  • Questions about what people are posting on X bring in Grok
  • API and MCP callers can send preset: "auto" to let Synero pick the preset and models per question; the API reports the pick as preset and routeReason
·v2.23feature
Web search turns on by itself when a question needs it
  • Ask about news, prices, the latest version of something, this year, or paste a link, and web search turns on automatically. The composer shows it in blue (Web · Auto) and one click turns it off
  • The same rule now applies to the API and the MCP tool: omit web_search and search runs only for time-sensitive questions, reported as webSearchAuto; send web_search: false to skip it
·v2.22improvement
Updated Terms and Privacy, new Refund Policy
  • The Terms of Service and Privacy Policy now cover Whop payments, the API, MCP and connectors, share links, private mode and the cookies we actually use
  • New Refund Policy at synero.ai/refunds: when we refund, and what a refund does to your subscription and credits
  • Clarified that unused credits carry over month to month while your subscription is active
·v2.21feature
Synero MCP in ChatGPT and Grok
  • MCP is now in the top menu, and synero.ai/mcp has step by step setup for ChatGPT (Plugins, Create custom MCP server) and Grok (grok.com custom connectors and Grok Bot plugins)
  • Both sign in with your Synero account over OAuth; Grok Bot can also use an API key header
  • Synero is listed in the official MCP Registry as ai.synero/council
·v2.20feature
Chat controls: regenerate, edit your last message, stop, copy
  • Regenerate the last answer, or edit your last message and rerun it, with the same 1 to 4 models you picked
  • A Stop generating pill appears while models answer, and Esc stops the run
  • Copy the whole chat in one click
  • Presets now set how many models the chips allow: Solo one, Duel two, Council and Deep up to four
  • The composer sends on Enter and adds a new line on Shift+Enter, like ChatGPT and Claude
·v2.19launch
A chat first app: pick 1 to 4 models, see where they agree
  • The app is now a chat. Pick any of GPT, Claude, Gemini and Grok in the message box with one click, or use a preset: Solo, Duel, Council or Deep
  • One model answers like a normal chat. With two or more, you get a consensus answer plus where they agree, where they differ, and an agreement score; each model's full answer is one tab away
  • Models now answer as themselves instead of playing personas, so the disagreements you see are real
  • The consensus answer starts under a second after the models finish and completes about four times faster. We tested 26 model and reasoning settings, including the new releases; GPT-5.6 Luna writes the consensus and GPT-6 Luna grades agreement
  • Stop a response mid-stream, and keep typing your next question while answers stream
  • New models: GPT-6.1 Sol, GPT-6 Luna, Claude Opus 5.5, Claude Sonnet 5.5 and Grok 4.7, 20 models in all. Council (balanced) now uses Claude Sonnet 5.5; Deep uses GPT-6.1 Sol, Claude Opus 5.5, Gemini 3.1 Pro and Grok 4.7; the fast tier uses GPT-6 Luna. Each costs the same or less than the model it replaces
  • API: SSE streams add a consensus event with the agreement analysis (usually during the synthesis, but it may arrive after complete, so read until the stream closes), and JSON responses add consensus. An explicit synthesizerModel of gpt-5.6-luna now runs with reasoning off
  • Fixes from a full code sweep: you can only rate questions you own or that are shared publicly; Stripe retries a paid invoice or subscription change it could not match to an account instead of dropping it; conversation history sent to the API is capped at the last 10 turns; checkout explains when an account has no email; and hundreds of garbled sentences in our guides now read correctly
  • A follow-up question no longer cancels the previous answer's agreement report, and the agreement grader's small cost is metered like every other model call
·v2.18fix
Auto now sizes every question you ask in the app
  • Auto now picks the council size for each question you ask in the app. Before, the app always sent your saved models, so every Auto question quietly ran the same medium council
  • A custom council runs exactly as you set it, including a synthesizer you picked or a full tier of advisors you chose by hand. Editing back to the default council switches you back to Auto
  • Templates on the top-tier models, like Due Diligence, Strategy Brief and Investment Analysis, keep their models and synthesizer after a reload instead of turning into Auto
  • Free trial questions and questions blocked by billing no longer wait on the router
  • API: with councilTier auto or instant, sending advisorModels or synthesizerModel runs a custom council with those models, and an empty advisorModels object no longer counts. The docs now say so, and the default config example is { "councilTier": "auto" }
·v2.17fix
Refusals and failed answers now say so
  • When a model declines a question (a safety filter refusal) or its stream fails, its advisor card now says why instead of showing a blank answer marked complete
  • If the synthesizer declines or fails, the synthesis panel shows the reason and the advisor answers stay on screen; the query still completes and saves
  • Provider rate limits and authentication failures now show their real reason on the advisor card instead of a generic message
  • API: synthesis-complete can carry an error message, and JSON responses include synthesisError when a requested synthesis failed. The MCP server hands your agent the advisor answers to synthesize itself
  • A synthesizer error that used to end the run with a fatal error now reports the same way, and the streaming MCP connector returns advisor answers it already has instead of failing the call, so an agent doesn't retry and pay for a second council
·v2.16improvement
A simpler council: Auto only, fewer pages, models over personas
  • Auto is now the only council mode. The Low, Medium, High picker, per-question override, and Alt+1/2/3 shortcuts are gone; hand-pick models under Advanced if you want a custom bench
  • The app nav is three items: Integrations (MCP connect plus API keys), Billing (now includes Usage), and History. Templates moved into the composer
  • Advisor chips and the consensus certificate now lead with the model name; the persona is a subtitle
  • One plan gate replaces the trial counter, the exhaustion screen, and the billing card, with one sentence and one button per state
  • The homepage leads with a real shared council where the models split, and the live try box is prefilled with a question that divides them
·v2.15improvement
Bring Your Own Keys retired
  • Every query now runs on Synero's platform keys. There is no longer a Settings page for adding your own OpenAI, Anthropic, Google, or xAI keys, and no per-provider model locking
  • Pricing is one plan: 3 free queries, then $10/month with $10 in credits included for all 15 models
  • API and MCP 402 responses no longer carry uncovered_providers or has_user_keys
·v2.14feature
GPT-6 Astra, Claude Fable 5.1 and Opus 5, Gemini 3.8 Flash, Grok 4.6
  • Eleven new models: GPT-6 Astra and the GPT-5.6 family (Sol, Terra, Luna) from OpenAI; Claude Fable 5.1 and Claude Opus 5 from Anthropic; Gemini 3.8 Flash, 3.5 Flash, and 3.5 Flash-Lite from Google; Grok 4.6 and 4.5 from xAI
  • Retired the previous generations (GPT-5.5 and older, Claude 4.x and Fable 5, Gemini 2.5 and 3 Flash Preview, Grok 4.20). Saved councils, presets, and API calls that name a retired model run on its current-lane replacement automatically. 15 models across 4 providers
  • Council tiers moved to the new generation: low runs GPT-5.6 Luna, Claude Haiku 4.5, Gemini 3.5 Flash-Lite, Grok 4.3; medium runs GPT-5.6 Terra, Claude Sonnet 5, Gemini 3.8 Flash, Grok 4.5; high runs GPT-5.6 Sol, Claude Opus 5, Gemini 3.1 Pro, Grok 4.6
  • Instant answers now run on Gemini 3.5 Flash-Lite. Built-in templates use the current flagship from each provider
  • Claude Sonnet 5 pricing corrected to Anthropic's current $2 / $10 per million tokens
·v2.13improvement
Faster, more reliable council calls from Claude and other agents
  • The query_council MCP tool now defaults to returning all four advisor responses plus guidance for your agent to synthesize them itself, roughly half the latency of a server-synthesized answer. Pass synthesis: true for the old single-answer behavior
  • API and MCP calls are now capped at medium tier and bounded by a server-side deadline, so a call can't run long enough to hit a client's tool timeout, no matter how it's configured
  • Fixed: council configs with custom advisor models or a custom synthesizer model set via the API could get silently overridden by tier defaults; explicit model choices are now respected
  • Fixed: an all-advisors-failed response could look like a successful, empty result to some MCP clients instead of a clear error
·v2.12launch
Auto mode, instant answers, image attachments, and a true conversation
  • Auto mode is the new default: a fast router reads each question and sizes the council to it, from a quick low-tier bench up to the flagship high tier. Every turn shows which council Auto picked and why
  • Instant answers: clearly trivial questions (arithmetic, dates, quick lookups) skip the council and get a direct answer from one fast model in about 2 seconds. One click runs the full council if you want the second opinion after all
  • Attach images: drop a screenshot, chart, or photo into your question and all four advisors see it. Up to 3 images per query, works over the API and MCP too
  • The council is now a conversation: questions and answers flow in one thread, with each advisor's full response a tap away behind its color-coded chip. The composer stays with you as you scroll
  • Advisor chips pulse while each model thinks, so you can watch the council deliberate in real time and expand any member mid-stream
·v2.11feature
Claude 5 family, GPT-5.4 Mini and Nano, smarter BYOK
  • Claude Fable 5, Claude Sonnet 5, and Claude Opus 4.8 are live. Anthropic's newest generation, available in any advisor seat or as the synthesizer
  • GPT-5.4 Mini and GPT-5.4 Nano join the catalog. The default council now runs GPT-5.4 Mini, Claude Sonnet 5, Gemini 3 Flash, and Grok 4.3
  • Retired xAI models (Grok 4, Grok 4.1 Fast, Grok 3 Mini) map automatically to Grok 4.3, so saved councils and API calls keep working. 22 models across 4 providers
  • BYOK is more transparent: advisors that ran on your own key show a badge, a key rejected by its provider is flagged Invalid in Settings, and OpenAI keys are checked for credits when you add them
  • Without a subscription, model pickers now focus on the providers you brought keys for, with a clear path to unlock the rest
·v2.10feature
Grok 4.3
  • Grok 4.3 is live on the xAI API. Always-on reasoning, 1M token context window, $1.25/$2.50 per million tokens. Slots into any advisor seat or as the synthesizer
  • 20 models across 4 providers now available
  • Available immediately across the dashboard, REST API, MCP server, and all agent framework integrations
·v2.8feature
Claude Opus 4.7 and Grok 4.20
  • Claude Opus 4.7 is live. Anthropic's new flagship model for coding, agents, and long-running workflows. Same $5/$25 per million tokens as Opus 4.6, with a new tokenizer and stronger reasoning
  • Grok 4.20 Reasoning and Grok 4.20 (non-reasoning) are live. xAI's newest flagship with a 2M context window and $2/$6 per million tokens
  • 19 models across 4 providers now available. Assign any model to any advisor slot, or choose any as your synthesizer
  • Available immediately on every surface: dashboard, REST API, MCP server, and all agent framework integrations
·v2.7feature
Bring Your Own Keys (BYOK)
  • Add your own AI provider API keys to use models for free. No subscription required
  • Supports OpenAI, Anthropic, Google, and xAI keys
  • Per-model billing: only pay credits for models using Synero's keys
  • Keys encrypted at rest with AES-256-GCM, never returned to the client
  • Key validation on save: tests your key against the provider before storing
  • Visual indicators showing which models are free (your key) vs paid (credits)
·v2.6feature
History Sidebar & Search
  • Persistent history sidebar on the council page. Browse and reload past queries without leaving the conversation
  • Full-text search across your query history
  • Date filtering: Today, This week, This month, All time
  • Click any past query to load it inline with all advisor responses and synthesis
  • Rename queries with custom titles
  • Infinite scroll in the sidebar for large histories
  • Collapsible sidebar with state persisted across sessions
·v2.5feature
Council Templates & Template Marketplace
  • 14 built-in council templates that transform advisors into specialized teams (Code Review, Due Diligence, Deep Research, and more)
  • Each template sets custom advisor personas, model assignments, and structured synthesis instructions
  • Template marketplace: create, publish, fork, and share templates with the community
  • Public template gallery at /templates with category filtering
  • Template support in the API and MCP server via the template parameter
  • Replaces the old prompt template system with a fundamentally more powerful approach
·v2.4feature
Context-Aware Council
  • Attach files to your council queries: upload documents, images, or code for advisors to analyze
  • Context field in the API and MCP server for programmatic file attachments
  • Updated API docs, code examples, and MCP tool schema to reflect new context parameter
  • JSON response format option for the public API
·v2.3feature
Agent Experience & MCP Server
  • Official MCP server: connect Synero as a tool in Claude Code, Cursor, and other MCP clients
  • Framework integration pages for LangChain, CrewAI, AutoGen, and Semantic Kernel
  • Interactive API playground for testing queries in the browser
  • JSON API mode for structured responses
·v2.2improvement
Next.js 16.2 & AllYourTech Feature
  • Upgraded to Next.js 16.2 and React 19.2.4
  • AllYourTech newsletter popup with dark theme
  • AllYourTech featured badge in site footer
·v2.1improvement
SEO Infrastructure & Content Expansion
  • 9 new workflow use-case pages and 4 category comparison pages
  • 4 new developer documentation pages
  • JSON-LD structured data, canonical URLs, and breadcrumb navigation across all pages
  • Open Graph images for social sharing
  • SEO dashboard updated with accurate coverage tracking
·v2.0feature
Positioning Overhaul & Trust Pages
  • New core messaging: "Your AI Council for questions that matter"
  • Homepage hero, features, and CTA updated to emphasize verification and cross-model consensus
  • Refreshed /vs/perplexity page: addresses Perplexity's Model Council launch with council-first differentiation
  • New /pricing page with transparent plan details, credit top-ups, and competitor comparison
  • New /security page covering data privacy, row-level security, encryption, and upstream provider policies
  • New /how-synero-works page with step-by-step methodology and verification rationale
  • 6 new workflow use-case pages: Verify ChatGPT Answers, Fact-Check AI Output, AI Second Opinion, Market Research, Product Strategy, and Architecture Decisions
  • Each use-case page uses the new decision-asset format: problem framing, real prompt example, agreement/disagreement analysis, and synthesis preview
  • Strengthened internal links across footer, hero CTAs, and cross-page navigation
·v1.7improvement
Collapsible Starter Prompts
  • Starter prompts now show category pills only. Click to expand templates, click again to collapse
  • Animated accordion transitions with staggered card reveals
  • Cleaner default view with less visual noise on the council page
·v1.6feature
OpenClaw Agent Skill
  • New /agent landing page for the OpenClaw agent skill
  • Install Synero as a terminal skill: openclaw install blovett80/synero
  • Links to ClawHub listing added to developers page and footer
·v1.5feature
Toggle-able Advisor Slots
  • Choose 1–4 active advisors per session using the new toggle controls in Configure
  • Single-advisor mode skips synthesis and shows the response full-width
  • Grid layout adapts dynamically to the number of active experts
  • Last remaining advisor cannot be disabled
·v1.4improvement
Mobile Navigation
  • Responsive hamburger menu for mobile and tablet screens
  • Full-screen overlay with backdrop blur and smooth transitions
  • Menu auto-closes on route change and prevents background scrolling
·v1.3feature
Platform Feature Drop
  • Comparison pages: see how Synero stacks up vs ChatGPT, Perplexity, Gemini, and more
  • Starter prompt templates across 5 categories: Research, Strategy, Creative, Technical, and Decision Making
  • Keyboard shortcuts: ⌘+Enter to submit, Escape to cancel, number keys 1-4 to expand advisors
  • PWA support: install Synero as a native-feeling app on any device
  • Light mode with full theme toggle support
  • Advisor ratings: rate individual responses and the synthesis
  • Advisor presets: save and load your favorite model configurations
·v1.2improvement
Models, SEO & Polish
  • Added GPT-5.4 as a selectable model across all advisor slots
  • 14 new SEO keyword landing pages targeting capability-based searches
  • Reorganized footer with responsive layout and categorized use-case links
  • Homepage updated to showcase all 14 available models
  • Use-case pages now display all 4 advisor responses
  • Fixed Stripe cancellation detection for subscriptions with cancel_at timestamps
  • Demo question stays visible while answers stream in
  • Admin dashboard shows accurate total user counts
·v1.1feature
Public API & Developer Tools
  • Public REST API with API key authentication
  • Developer landing page with interactive API documentation
  • 25 persona-based use-case pages with full SEO metadata
  • API usage statistics and auth method tracking in admin dashboard
  • Authentication now required for all API queries
  • Improved app navbar with avatar menu and mobile responsiveness
·v1.0.1fix
Launch Week Polish
  • Google Analytics 4 with signup conversion tracking
  • FAQ section added to the landing page
  • GitHub-Flavored Markdown table rendering in advisor output
  • Updated model references to feature GPT-5.2 and Claude Opus
  • Fixed orbital visualization alignment on the homepage
·v1.0launch
Synero Launch
  • Multi-model AI council: ask one question, get four perspectives synthesized into one answer
  • Four contrasting advisor roles: Architect, Philosopher, Explorer, and Maverick
  • 14 AI models from OpenAI, Anthropic, Google, and xAI
  • Stripe billing with free tier and Pro subscription
  • Supabase authentication with email and social sign-in
  • Animated demo walkthrough on the landing page
  • Monthly credit grant of $10 for all accounts