📊 ChartSmith MCP Server

A high-performance chart generation MCP server — built for speed, powered by AI, production-ready.

🚀 First-ever comprehensive chart generation MCP server - Turn your SQL results into stunning visualizations with zero configuration and maximum performance.

🎯 Why ChartGenius?

💡 Inspired by Industry Leaders

• Hex.tech's $70M success proves chart generation is core business infrastructure
• Built to solve real bottlenecks like O(n²) algorithms and blocking operations
• MCP-native design for seamless integration with any client

⚡ Performance First

• Optimized O(n) data paths
• Concurrent rendering with guardrails (timeouts, semaphores, input caps)
• Intelligent caching (memory/Redis)

🧠 AI-Powered Intelligence

• Smart chart type detection based on data patterns
• Natural language to visualization conversion
• Automatic insight generation for business intelligence
• Accessibility optimization for WCAG compliance

🚀 Quick Start

Installation

pip install chart-genius-mcp  # package name remains; server id: chartsmith

Docker Quick Start

docker build -t chartsmith-mcp .
docker run --rm -it chartsmith-mcp                 # stdio
docker run --rm -it -p 8000:8000 -e RUN_HTTP=1 chartsmith-mcp  # http

Using docker-compose:

docker compose up -d mcp-stdio
docker compose up -d mcp-http

🔌 Transports

• Stdio (MCP stdio)
• HTTP JSON (POST /mcp) and SSE (POST /sse)
• Health: /health, Readiness: /ready, Metrics: /metrics

⚙️ Environment

• CHART_MAX_CONCURRENCY (default 8)
• CHART_TOOL_TIMEOUT_MS (default 60000)
• CHART_TOOL_TIMEOUT_MS_IMAGES (image timeouts)
• CHART_TOOL_TIMEOUT_MS_HEAVY (batch/dashboard)
• CHART_MAX_ROWS (input size guard)
• DISABLED_TOOLS (comma-separated)
• AI: OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, CHART_AI_PROVIDER

🧰 Tools

• Per-chart: bar, line, scatter, histogram, area, box, violin, bubble, pie, heatmap, treemap, sunburst, sankey, choropleth
• Full-auto: generate_chart_auto
• Deprecated: generate_chart
• Cache: manage_cache (stats/clear/optimize)
• Perf: get_performance_stats
• Prompts/Resources: list_prompts/get_prompt, list_resource_templates/read_resource

📦 Resources

• Datasets: dataset:sample/sales_small, .../scatter_small, .../heatmap_pivot, .../choropleth_iso3
• Themes: theme:modern, theme:corporate, theme:dark, theme:accessible

🧪 Tests & Benchmarks

• Pytest suite: tools, engines, formats, AI (mock/live), cache, guardrails, MCP stdio/HTTP
• Benchmark harness: p50/p90 latency script and smoke tests

📈 Prometheus Metrics

• Endpoint: /metrics (Prometheus exposition format)
• Metrics include: request counts, tool call totals, tool errors, and duration histograms.
• Example scrape config:

scrape_configs:
  - job_name: chartsmith
    static_configs:
      - targets: ['localhost:8000']

🔐 Security & Limits

• Input size caps, timeouts, concurrency limits
• Tool filtering via env

🧭 Smithery

• Stdio: configure Smithery to invoke python -m chart_genius_mcp with server id chartsmith
• HTTP: point Smithery at the container’s /mcp endpoint (port 8000)

📦 Docker Images (CI)

• GitHub Action builds and publishes multi-arch images to GHCR on tags v*.
• Image: ghcr.io/<owner>/<repo>:<tag> and :latest.
• Example pull:

docker pull ghcr.io/<owner>/<repo>:latest