SlashMCP

MCP-powered AI workspace for document intelligence, multi-agent orchestration, and dynamic MCP server management.

Live Demo: slashmcp.vercel.app | Repository: github.com/mcpmessenger/slashmcp

• 🚀 Dynamic MCP Server Registry - Manage and invoke Model Context Protocol servers from the chat interface
• 📄 Document Intelligence - Drag-and-drop uploads with OCR via AWS Textract and vision analysis via GPT‑4o
• 📊 Financial Data - Stock quotes, charts, and prediction market data through integrated MCP servers
• 🤖 Multi-Agent Framework - Orchestrate multiple AI agents with standardized MCP tool access
• 🔄 LLM Switching - Seamlessly switch between OpenAI, Anthropic, and Gemini providers at runtime

Project repo: https://github.com/mcpmessenger/slashmcp

---

At a Glance

• Frontend: Vite + React + TypeScript + Tailwind + shadcn/ui
• State & Data: TanStack Query, Supabase client
• Workers / Integrations: Supabase Edge Functions (uploads, vision-worker, textract-worker, chat, job-status)
• Cloud Services: AWS S3 + Textract, OpenAI GPT‑4o, Google Gemini (fallback), Supabase Postgres
• Storage: S3 bucket arn:aws:s3:::tubbyai-products-catalog
• Environment management: .env for Vite, Supabase function secrets for service keys

---

Getting Started

Prerequisites

• Node.js ≥ 18.18 (recommend using nvm)
• npm ≥ 9 or pnpm/bun if you prefer (project is currently npm locked)
• Supabase CLI (npm install -g supabase or npx supabase)
• AWS credentials with access to Textract + S3 bucket above
• OpenAI API key (GPT‑4o vision)
• Optional: Google Generative AI key (for Gemini fallback), Whisper / Google Cloud TTS keys for upcoming features

Setup

git clone https://github.com/mcpmessenger/slashmcp.git
cd slashmcp
npm install

Create .env.local (Vite automatically loads VITE_* variables):

Quick Setup: Copy .env.local.example to .env.local and fill in your values.

Manual Setup: Create .env.local with:

VITE_SUPABASE_URL=https://<your-supabase-ref>.supabase.co
VITE_SUPABASE_FUNCTIONS_URL=https://<your-supabase-ref>.supabase.co/functions/v1
VITE_SUPABASE_PUBLISHABLE_KEY=<supabase-anon-key>
VITE_ALPHA_VANTAGE_API_KEY=<alpha-vantage-key>
VITE_MCP_GATEWAY_URL=http://localhost:8989/invoke # MCP gateway proxy endpoint
VITE_SUPABASE_REDIRECT_URL=http://localhost:8080          # OAuth redirect URL (local dev)
# For production: Set VITE_SUPABASE_REDIRECT_URL to your production domain in Vercel env vars
TWELVEDATA_API_KEY=<twelve-data-key>                # optional fallback provider
ALPHAVANTAGE_CACHE_TTL_MS=300000 # optional Supabase function cache TTL (5 min)
POLYMARKET_CACHE_TTL_MS=120000   # optional cache window for Polymarket lookups

Note: See LOCAL_SETUP.md for detailed local development setup, including S3 CORS configuration for file uploads.

Configure Supabase Edge Function secrets (one-time per project):

npx supabase secrets set \
  --project-ref <supabase-ref> \
  PROJECT_URL=https://<your-supabase-ref>.supabase.co \
  SERVICE_ROLE_KEY=<supabase-service-role-key> \
  OPENAI_API_KEY=<openai-key> \
  GEMINI_API_KEY=<optional-gemini-key> \
  GEMINI_IMAGE_MODEL=gemini-2.5-flash-image \
  ALPHAVANTAGE_API_KEY=<alpha-vantage-key> \
  TWELVEDATA_API_KEY=<twelve-data-key> \
  ALPHAVANTAGE_CACHE_TTL_MS=300000 \
  POLYMARKET_CACHE_TTL_MS=120000 \
  GOOGLE_CLIENT_ID=<google-oauth-client-id> \
  GOOGLE_CLIENT_SECRET=<google-oauth-client-secret> \
  AWS_REGION=<aws-region> \
  AWS_ACCESS_KEY_ID=<aws-access-key> \
  AWS_SECRET_ACCESS_KEY=<aws-secret> \
  AWS_SESSION_TOKEN=<optional-session-token> \
  AWS_S3_BUCKET=tubbyai-products-catalog

Serve the dev app alongside Supabase functions:

npm run dev             # Vite dev server
npx supabase functions serve --env-file supabase/.env # optional local function emulation

---

Key Features & Flow

1. Upload & Job Tracking

   • src/components/ui/chat-input.tsx drives the menu actions (Upload Files, Document Analysis, Image OCR, Voice Assistant).
   • Frontend calls supabase/functions/uploads to create a processing_jobs row and receive an S3 presigned PUT URL.
   • Upload progress is surfaced in the UI; processing_jobs status transitions are polled via job-status function.

2. Vision Analysis

   • vision-worker fetches the uploaded asset from S3 via presigned GET, submits to GPT‑4o (with JSON-structured prompt), and persists results to analysis_results.
   • Gemini fallback helpers exist and can be toggled back in if desired.

3. Textract OCR

   • textract-worker uses AWS Textract (sync for images, async for PDFs) and stores the OCR text + raw response.
   • Frontend handles Textract errors gracefully (e.g., “No text detected”) and falls back to vision summary.

4. Supabase Schema

   • Migration supabase/migrations/20250108120000_create_processing_jobs.sql provisions:
     • processing_jobs: metadata for each upload
     • analysis_results: OCR + vision outputs (ensure the migration is run on the Supabase project)
     • analysis_results_job_id_unique unique index is required for onConflict: "job_id" upserts

5. Future Scaffolding

   • Hooks and menu options exist for Whisper ASR and Google TTS pipelines.
   • Natural language prompts (or /imagine <prompt>) funnel through supabase/functions/image-generator, which now targets the Gemini 2.5 Flash Image ("Nano Banana") model for generation.

6. Stock Quotes & Charts

   • Enter /quote AAPL, /stock MSFT 3m, or /ticker NVDA 1y in the chat input to fetch Alpha Vantage data.
   • The assistant renders a price card with daily trend chart, change metrics, and key stats.
   • MCP endpoint support is scaffolded so /alphavantage-mcp get_stock_chart symbol=NVDA uses the same rendering pipeline (Supabase edge function mcp).
   • Quotes are cached for 5 minutes and automatically fall back to Twelve Data (if TWELVEDATA_API_KEY is set) when Alpha Vantage hits premium/rate limits.

7. Prediction Markets

   • Use /polymarket-mcp get_market_price market_id=us_election_2024 to pull live odds from Polymarket via the Supabase mcp function.
   • Responses include best bid/ask, implied probability, liquidity, and are cached for 2 minutes (configurable via POLYMARKET_CACHE_TTL_MS).

8. Provider Switching

   • Use /model openai, /model anthropic, or /model gemini to switch the backing LLM at runtime (defaults to OpenAI).
   • Environment variables and Supabase function secrets (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY) must be configured for each provider you plan to use.

9. MCP Slash Commands & Registry

   • Authenticate via the header sign-in button (Google OAuth) or the fallback chat command /slashmcp login [email protected] password=secret.
   • Once signed in, manage servers with /slashmcp list, /slashmcp add <name> <https://gateway> (plus optional auth=/key= parameters), and /slashmcp remove <name|serverId>.
   • Provider shortcuts like /gemini or /playwright map to pre-defined presets and prompt for any required secrets.
   • Natural language detectors route common stock and Polymarket questions to the appropriate MCP server automatically when available.
   • Dynamic registry records are stored per-user in Supabase (mcp_servers table) and proxied through new edge functions (mcp-register, mcp-get-registry, mcp-remove, mcp-proxy).

---

Available Scripts

Command	Description
npm run dev	Start Vite dev server
npm run build	Production build
npm run build:dev	Development-mode build (useful for staging)
npm run preview	Preview built assets
npm run lint	ESLint across the project

Supabase Edge Functions can be deployed with:

npx supabase functions deploy <function-name> --project-ref <ref>

---

Directory Highlights

• src/components — Landing page and UI components (Hero, Features, Architecture, CTA).
• src/components/ui — shadcn/ui wrappers.
• src/lib/api.ts — Frontend API client (uploads, Textract, vision).
• supabase/functions/* — Edge functions for uploads, job status, Textract, GPT vision, and chat.
• supabase/functions/_shared/database.types.ts — Generated Supabase typings shared across functions.
• supabase/migrations — Database schema for job tracking + analysis results.

---

Deployment Notes

• Ensure Supabase migrations have been applied (analysis_results vision columns and unique index).
• Set S3 CORS to allow your dev/prod origins and methods (GET,PUT,POST,HEAD).
• When deploying to new environments, rotate and populate Supabase function secrets before invoking uploads/vision workers.
• Voice synthesis (Whisper, Google TTS) still requires additional secrets; Gemini image generation expects GEMINI_API_KEY (and optional GEMINI_IMAGE_MODEL if you override the default gemini-2.5-flash-image). Favicon and logo assets now live under public/Untitled design.svg and public/Untitled design (12/14).png.
• MCP gateway should expose a single /invoke endpoint compatible with the JSON payload emitted by src/lib/mcp/client.ts.

---

Security & Maintenance

• Run npm audit (or npm audit --omit=dev) regularly; see “Security Scan” section below.
• Restrict Supabase service-role key usage to server-side contexts only; never expose it to the browser.
• Use AWS IAM roles with least-privilege access to Textract and the specific S3 bucket.
• Consider setting up Supabase Row Level Security policies if multi-tenant support is required.

---

Roadmap / TODO

• Wire up Whisper transcription pipeline with Supabase edge worker + queue status updates.
• Integrate Google Cloud Text-to-Speech for voice responses.
• Re-enable Gemini fallback with configurable provider selection and cost tracking.
• Add automated testing (Cypress/Playwright) for upload + analysis flow.
• Harden Supabase RLS and add auth once user accounts are introduced.

---

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Built with ❤️ by the SlashMCP team