Build history

Direct SDK connections to OpenAI, Anthropic Claude, and Google Gemini. Unified streaming responses with provider-specific optimizations.

openai, @anthropic-ai/sdk, @google/generative-ai packages with unified response adapter layer

Enhanced model picker with capability indicators, context length display, and task-specific recommendations.

Model metadata system, capability tags (vision, code, function-calling), dynamic recommendations

Real-time provider status with automatic failover to healthy providers when issues are detected.

Health check endpoints, circuit breaker pattern, graceful degradation

Faster streaming with reduced time-to-first-token across all providers. Consistent streaming behavior regardless of provider.

ReadableStream with backpressure handling, provider-agnostic chunk processing

Accurate token counting per provider using official tokenizers. Pre-chat token estimation for better model selection.

tiktoken for OpenAI, claude-tokenizer approximation, vertex token estimation

Smart retry logic with exponential backoff and provider-specific error handling. Clear error messages with actionable next steps.

AI-powered chat assistant trained on ARKA AI documentation. Get instant answers about features, pricing, troubleshooting, and more.

RAG-based AI with knowledge base embeddings, streaming responses, conversation persistence via sessionStorage

Create and track support tickets for personalized assistance. View ticket history, status updates, and agent replies.

PostgreSQL tickets/messages tables, real-time polling, email notifications via Resend

Attach images, documents, and logs to tickets and replies. Supports common file types with image previews.

S3 file storage with presigned URLs, JSONB attachments field, 24-hour URL expiry

Tickets automatically refresh every 15 seconds. 'Updated' indicator appears when new messages arrive.

Custom usePolling hook with tab visibility detection, timestamp comparison for change detection

Clean navigation with Overview, Chat, and Tickets tabs. Chat opens as a popup from the tab.

Response templates for support agents. Quick-insert via dropdown or keyboard shortcuts (e.g., /billing).

CRUD API, category grouping, usage tracking, real-time shortcut detection

Manage AI knowledge articles for accurate support responses. Search, categorize, and reindex embeddings.

Chat messages now properly render markdown links [text](url) and bold **text** formatting.

Create projects to organize your conversations and images. Each project can have a custom name, emoji icon, and color.

PostgreSQL projects table, workspace-scoped project management, real-time sidebar updates

Easily move conversations between projects from the conversation header. One-click project assignment.

Conversation-project relationship, optimistic UI updates, workspace access validation

Personalize projects with emoji icons and 9 color options: Gray, Red, Orange, Yellow, Green, Teal, Blue, Purple, and Pink.

Projects appear in the sidebar for quick filtering. Click a project to see only its conversations.

Generate images from text prompts using DALL-E 3 and Nano Banana Pro. ARKAbrain automatically routes to the best model for your prompt.

OpenAI DALL-E 3 API, OpenRouter Nano Banana Pro, S3 image storage with presigned URLs

Upload images or use generated ones as a base. Edit with AI using natural language instructions or reimagine with new styles.

Image-to-image pipeline, mask-free editing, style transfer with preset options

ARKAbrain analyzes your prompt to choose the optimal model. DALL-E 3 for photorealistic, Nano Banana Pro for artistic styles.

Prompt analysis for style detection, capability matching, confidence-based routing

One-click style presets: Ghibli, Pixar, Anime, Oil Painting, Watercolor, Cyberpunk, and more.

Plugin cards now fully responsive on mobile with stacked layouts and adjusted spacing.

View complete history of all Quick Tools runs in your workspace. See inputs, outputs, model used, and timestamps.

Persistent tool_runs table with workspace scoping, paginated history view

Re-run any previous tool with the same input. Perfect for iterating on results or trying different models.

Slide-out panel shows recent runs while you work. Click any entry to view details or re-run.

Create team organizations with dedicated workspaces. Invite members and collaborate on AI projects together.

Share conversations with your team. Team members can view shared conversations while the owner retains full control.

Invite team members via email, assign roles (Owner, Admin, Member, Viewer), and manage permissions from the settings page.

Simple seat-based pricing. $99/month includes 5 seats, additional seats at $15/month each. All team members get Pro access.

Four roles with different permissions: Owner (full control + billing), Admin (manage members), Member (full usage), Viewer (read-only).

5 comprehensive docs pages covering URL audits, AI content generation, analytics insights, and GitHub code changes.

5 docs pages for Gmail, Drive, Sheets, and Docs integrations with AI generation guides.

Email inbox with AI classification (Important, Promotional, Newsletters, Transactional), draft generation, and scheduled sending.

Gmail API with OAuth2, AI classification pipeline, draft/scheduled queue system

File browser with search, AI document analysis, and Q&A discussions on file contents.

Drive API v3, document parsing (Docs, Sheets, PDF), vector-based Q&A

Read/write spreadsheet data with AI content generation from natural language prompts.

Sheets API v4, A1 notation support, batch read/write operations

Document creation and modification with AI writing assistance and style suggestions.

Docs API v1, section-based editing, formatting preservation

Connect your GSC properties to view clicks, impressions, CTR, and position data with trend analysis.

Search Console API, OAuth2 flow, data sync with 7/28/90 day ranges

Landing page performance, sessions, bounce rate, and engagement metrics from GA4.

GA4 Data API, property selection, landing page dimensions

Automatic detection of optimization opportunities: Low CTR pages, Quick Wins (position 8-20), Click Drops, and Hidden Gems.

Rule-based insight detection, severity scoring, action recommendations

Connect your GitHub repositories for automated SEO code analysis.

GitHub OAuth App, repository listing, branch detection, file tree scanning

Automatic detection of SEO issues in your code: missing meta tags, Open Graph, schema markup, semantic HTML.

AST parsing for HTML/JSX, pattern matching for meta tags, JSON-LD validation

Generate PRs with SEO fixes directly from approved suggestions. Includes side-by-side diff viewer.

GitHub API for PR creation, unified diff generation, batch commit support

6 content types: Guide, How-To, Comparison, Listicle, Case Study, FAQ. Each with optimized templates.

Multi-step generation: outline → review → full content. Token-optimized prompts per content type

5 voice options (Professional, Casual, Technical, Friendly, Authoritative), 3 reading levels, word count control.

Prompt engineering with tone modifiers, audience-aware vocabulary adjustment

Monthly/weekly/agenda views for content planning with keyword association and draft linking.

Calendar state management, drag-and-drop scheduling, status workflow (Planned → Draft → Published)

Link GSC insights directly to content drafts when creating new content.

25+ SEO rules across 6 categories: Meta Tags, Headings, Images, Links, Schema, Performance.

Puppeteer-based crawling, Cheerio HTML parsing, rule engine with severity scoring

Crawl single pages or up to 25 pages following internal links from the starting URL.

BFS link following, robots.txt respect, concurrent request limiting

Organize audits into projects. Track URLs, audit history, and improvement progress over time.

Project CRUD with Drizzle ORM, audit job queue, progress tracking

Subscribe to individual plugins with tiered pricing. Solo, Pro, and Agency tiers available.

Stripe subscription integration, usage metering, feature flags

Attach files to chat messages for context-aware AI responses. Pro feature.

S3 file storage, presigned URLs, file type validation, size limits

Extract text from PDF documents for AI analysis. Handles text-based PDFs.

unpdf library for serverless compatibility, fallback handling for image PDFs

Visual upload progress, drag-and-drop support, error display with retry.

Multiple fixes for PDF extraction in Vercel serverless environment.

Multi-factor model selection based on: task type, complexity, required capabilities, and cost efficiency.

Scoring algorithm with weighted factors, confidence calculation, fallback chains

Automatic prompt optimization for better AI responses without changing user intent.

Context injection, instruction clarification, format specification

See estimated savings vs. direct API usage with model-by-model comparison.

Token cost calculation, subscription amortization, usage tracking

Pre-built templates for common tasks with one-click population.

'Fast' mode now actually prioritizes speed and cost over capability.

Free: 25/day, Starter: 100/day, Pro: 500/day with clear usage display.

Redis-based counter with daily reset, middleware enforcement

Explore ARKA-AI without payment. Limited daily messages, no file uploads.

Plan hierarchy: free → starter → pro, feature flags per plan

Clearer, actionable error messages when API calls fail or limits are reached.

Workspace plan now properly syncs when subscription changes.

Updated Gemini models to latest OpenRouter slugs, removed deprecated gpt-4-turbo.

Automatic AI model selection based on task analysis. Access GPT-4, Claude, Gemini, and 50+ models.

Task classification, capability matching, confidence-based routing

Use your own API keys for OpenAI, Anthropic, OpenRouter. Pay providers directly with no markup.

Encrypted key storage, per-request key injection, provider abstraction

Try ARKA-AI with email verification. 3 free daily tool runs before signing up.

Stateless email tokens, usage tracking, upgrade prompts

Comprehensive docs with guides, API reference, and code examples.

Compare outputs from different AI models side-by-side on the same prompt.

Rate responses and report issues directly from the chat interface.