Multi AI models. Adaptive memory. Document intelligence. Image and video generation. Web search with citations. Cost dashboards. And an optimization engine that makes it all cheaper.
Start FreeStop juggling ChatGPT, Claude, and Grok in separate tabs. Selina gives you all of them in a single thread.
GPT-5.4, Claude Opus, Grok, and more — OpenAI, Anthropic, and xAI all accessible from one chat.
Start with GPT for code, switch to Claude for writing, use Grok for an image — all in one thread.
Models organized by provider with tier indicators. Always know exactly what you are choosing.
Every response shows what it cost and how much was saved. No hidden charges.
See exactly which model handled each response — requested vs actually used.
Document chunks, topic primer, routing info, latency, and savings % — all visible inline.
Every conversation is saved. Come back tomorrow or next month — it is all there.
Conversations are automatically titled after the first exchange. No manual naming needed.
Every conversation teaches Selina what you care about. Conversation #50 is smarter than #1.
Selina extracts topics from every chat and builds a personal knowledge map — per project, per user.
After 3+ interactions on a topic, Selina builds a summary that primes future conversations.
Knowledge summaries refresh every 5 interactions, getting more accurate over time.
Mention your book project in a new conversation and Selina already knows the themes and goals.
Selina recognizes related topics even when phrased differently. No exact wording required.
View, explore, and delete your learned topics. See interaction counts and sample queries.
Click any topic to see its full knowledge summary, recent queries, and interaction history.
Visual heatmap of all your interactions, color-coded by task type. See patterns at a glance.
See your most reused cache entries and how much they have saved through repeated hits.
Bar chart of memory units, tokens, and cost organized by task type (QA, code, creative, etc.).
Topics in your "Book Project" don't contaminate "Client Marketing." Intelligence stays organized.
After responses, contextual messages show topic recognition, cache hits, and savings milestones.
Sort topics by most recent, most interactions, name, or cost. Paginated for large knowledge bases.
Toggle web search on any conversation. Get sourced, cited answers — not hallucinations.
Search the web from inside any conversation. One toggle to activate.
Web search enabled on GPT-5.4, GPT-4o, and other OpenAI models.
Web search enabled on Claude Opus, Sonnet, and Haiku.
Every claim is sourced. Clickable links so you can verify anything the model says.
Strict instructions ensure the model only reports facts from search results. No fabricated data.
The model knows today's date and won't confuse last year's results with current information.
Enable your location and "near me" queries, weather, and local recommendations just work.
Blue and amber indicators on each message show web search status and results.
Your documents become part of every conversation. Selina indexes everything and surfaces what matters.
PDF, DOCX, TXT, Markdown, CSV, Excel, PowerPoint, PNG, JPG — all automatically indexed.
Text is extracted from every document type automatically. No manual copy-paste.
Documents are split into searchable chunks with overlap for seamless context boundaries.
Every chunk is embedded for semantic search. Find information by meaning, not just keywords.
Toggle "Docs" and Selina pulls the most relevant sections from your knowledge base.
The most relevant document sections are injected into the conversation as context for the model.
Document search respects project boundaries. Only the active project's documents are queried.
Upload a new version and it auto-links as v2, v3, etc. Full version history preserved.
Selina warns you if you upload a file that already exists. No wasted storage.
Rename any document inline. Titles are editable without re-uploading.
Reassign documents to different projects from the context menu.
Filter your knowledge base by PDF, DOCX, CSV, and other file types.
View only documents assigned to a specific project or unassigned (General).
Sort documents by newest, oldest, or name (A-Z / Z-A).
See total documents, total chunks, and total tokens embedded at a glance.
Generate visual and written content directly inside your conversations.
Standard quality image generation directly in chat.
High quality image generation for more detailed and polished results.
OpenAI-powered image generation as an alternative style option.
5-second video clips via Grok Imagine Video. Videos play directly in your conversation.
Generate formatted PDF documents from conversation context.
Generate Word documents from conversation context.
Images from the conversation are automatically embedded into generated documents.
Generated documents are automatically added to your knowledge base for future reference.
Draft emails with subject lines and structured body text.
One-click mailto links to open drafted emails in your email client.
Email HTML is formatted to render correctly in Outlook, Gmail, and other clients.
Upload images, screenshots, or diagrams and ask questions about them.
Images go to vision models. Small PDFs process inline. Large docs auto-add to knowledge base.
Attach multiple files to a single message. Each is classified and routed automatically.
Color-coded projects, persistent conversations, and a sidebar that keeps up with your work.
Create projects with custom names, descriptions, and 10 color options.
Add context to each project so you remember what it is for.
Assign documents to projects. Search respects project boundaries.
Assign conversations to projects. Filter chat history by project.
Topic knowledge is scoped per project. No cross-project contamination.
Expand any project to see its documents and conversations. Jump to chat in one click.
Start a new conversation inside a project context directly from the Projects page.
Deleting a project moves all items to General. Nothing is lost.
View conversation history scoped by project. Find exactly what you need fast.
Clean up your sidebar without deleting. Archived conversations are always recoverable.
Full transparency on every message. Dashboards that show exactly where your money goes.
Interactive area chart showing what you paid vs what was saved over time.
See spending and savings broken down by day across any period.
See how much was saved by smart routing, caching, compression, and document optimization.
Which models you use most, request counts, and cost per model.
Full table of every request with model, tokens, cost, savings, and latency.
Per-request optimization breakdown showing exactly which layers saved you money.
Clear comparison on every period. Real numbers, not estimates.
Dashboard breaks down usage, cost, and savings by individual API key.
7 days, 30 days, 90 days, or all time on every dashboard page.
Average latency, smart routes executed, tokens in, and tokens saved.
See what percentage of your requests are served from cache.
Every request passes through the SelinaFOLD engine. You don't configure anything — it just works.
Simple questions route to efficient models automatically. Up to 95% savings.
Each query is classified by type — trivial, lookup, creative, code, reasoning, complex.
Identical requests return instantly from cache. Sub-50ms response time.
Same question in different conversations still gets a cache hit.
Similar questions get cached responses. Ask the same thing differently and still save.
A cached response from a budget model won't be served to a premium model request.
Whitespace, redundant content, and formatting noise are cleaned automatically.
JSON content in prompts is minified to reduce token count.
Comments in code blocks are stripped before sending to the model.
Large CSV data is summarized to header + sample rows instead of sending thousands of rows.
Long conversations are intelligently summarized so the model focuses on what matters.
Conversation summaries are cached so the same conversation doesn't re-summarize.
Only the most relevant document sections are injected. No context window waste.
Low-relevance document chunks are dropped before reaching the model.
Time-sensitive queries get fresh results. Stale cache entries expire faster.
Repeated patterns (summarize, translate, explain) get progressively faster cache hits.
Automatic prompt caching at the Anthropic and OpenAI provider layer.
Prompts are pre-screened to prevent provider rejections and wasted fees.
Cache expiration times vary by query type — temporal, search, and standard get different TTLs.
Three plans. Self-service everything. No support tickets needed.
Starter ($29/mo), Pro ($79/mo), Power ($249/mo). Pick what fits.
$10, $25, $50, or $100. Use them whenever — no expiration.
Upgrade or downgrade directly from the app. No support request required.
Cancel your subscription anytime from inside the app. No hoops to jump through.
Manage payment methods, view invoices, and update billing info directly.
Full log of every subscription payment and top-up purchase with status.
All payments processed by Stripe. Selina never sees your card number.
Your balance is shown throughout the app so you are never surprised.
Adjustable text, real-time progress, mobile-ready — built for how people actually use AI.
Use +/- controls to resize chat text. Preference saved across sessions.
Watch stages as they happen: optimizing, searching docs, loading memory, generating.
Fresh responses stream in word-by-word. Cache hits display instantly.
Collapsible sidebar that remembers its state. Full-width chat when collapsed.
Slide-out navigation on mobile devices. Full feature access on any screen.
Floating button for quick navigation in long conversations.
Easy on the eyes for long work sessions. Every page, every component.
Set your display name and location. Selina uses them to personalize your experience.
Hit $1 or $5 saved in a session? Selina celebrates with you.
Every piece of data is scoped to your account. Zero cross-user access by design.
All caches, documents, memory, and topics are scoped to your account.
Cache entries include your user ID. No one else sees your cached responses.
Sign up with Google for one-click access. No password to remember.
Traditional registration with bcrypt-hashed passwords.
Secure reset flow with expiring tokens sent via SendGrid.
Per-second burst protection prevents abuse of your account.
Per-minute limits add a second layer of protection.
Prompts are screened before reaching providers to prevent TOS violations.
Enable for better local answers. Disable anytime. Never shared externally.
Registration rejects known disposable email domains automatically. No throwaway accounts.
IP-based rate limits prevent bulk account creation from a single source.
Jailbreak and role-confusion attempts are detected and blocked before reaching the model.
Repeated failed login attempts trigger a lockout to protect your account.
features in one platform
One account. Multi AI models. Intelligence that grows with you.