ai-netcafe

Hosted open-source AI apps, callable by your agent over MCP — and from your terminal.

npx ai-netcafe compare "dedupe a JS array preserving order"

MODEL              COST       LATENCY
deepseek-v4-flash  $0.00012    7.6s
gemini-3.5-flash   $0.000341   10.0s
GLM5.2             $0.001828   9.0s

cheapest: deepseek-v4-flash   fastest: deepseek-v4-flash
15.2× between cheapest and priciest

The point: costs are metered, not quoted

Every number above is read from the upstream usage response, not derived from a published price list. That distinction matters more than it sounds — while building this we found the obvious conversion formula was wrong for several models (one was off by ~9×, another by ~3×), and only billing-delta measurement matched reality. Full method and data.

Connect it to an agent

npx ai-netcafe install     # prints config for Claude Code / Desktop / Cursor

Or directly:

claude mcp add --transport http ai-netcafe https://ainetcafe.com/mcp

Skills the agent gets

Skill	What it does
compare_models	One prompt across many LLMs, with real metered cost + latency
translate_pdf	Translate a PDF keeping formulas, figures and two-column layout intact
deep_research	Plans sub-questions, searches, reads sources, writes a cited report
make_slides	Generates a real downloadable .pptx
list_apps / get_app	Catalogue of the hosted open-source apps, with source repos
ask_model / list_models	Single model call; model list with prices

Backed by open-source projects — PDFMathTranslate, GPT Academic, GPT Researcher, NextChat, LobeChat — hosted pre-configured so you can try them without installing Docker, Python, or supplying an API key.

CLI commands

npx ai-netcafe compare "<prompt>"   # multi-model comparison with real costs
npx ai-netcafe ask "<prompt>"       # single model
npx ai-netcafe apps                 # hosted apps + source repos
npx ai-netcafe models               # models with per-million prices
npx ai-netcafe install              # MCP config for your agent

Quota

Anonymous calls get a small free quota — enough to try. Heavy skills (whole-PDF translation, deep research, slide generation) need an account or your own key, because one of those runs costs dozens to hundreds of model calls.

export ALLROUTER_API_KEY=sk-...   # unlimited, billed to you

Links

• Docs: https://ainetcafe.com/mcp.html
• Live cost comparison (updated daily): https://ainetcafe.com/lab/llm-real-cost-comparison
• A2A agent card: https://ainetcafe.com/.well-known/agent-card.json

MIT