--- name: t2000-services description: >- Discover MPP services payable via `t2 pay`. Use when the user asks "what can I pay for?", "what AI models are available?", "show me the service catalog", "is there a weather API?", or any other discovery question. Pairs with the t2000-pay skill (discovery first, then pay). license: MIT metadata: author: t2000 version: "1.0" requires: t2000 CLI (npm install -g @t2000/cli) --- # t2000: Discover MPP Services ## Purpose Browse the live MPP gateway catalog at `mpp.t2000.ai` to find a service that matches the user's intent (chat, image gen, search, weather, email, code exec, mail, etc.) before calling `t2 pay`. The catalog has 40+ services across 88 endpoints with prices ranging from $0.005 to $30 per call. ## Rules 1. **Discover before paying.** Don't guess a URL — call `t2 services search` (CLI) or `t2000_services` (MCP) first. Service paths + pricing change as the gateway expands. 2. **Pick the cheapest endpoint that satisfies the user.** Many services have multiple tiers (e.g. `openai/v1/chat/completions` at $0.01 vs `openai/v1/audio/speech` at $0.05). Surface options. 3. **Surface pricing to the user before signing.** Every `t2 pay` write is opt-in via the user's own keypair — they deserve to know what they're spending. 4. **Live source of truth.** The catalog is fetched live from `https://mpp.t2000.ai/api/services` — what shows up via `t2 services search` is exactly what `t2 pay` can talk to. ## Commands ```bash # Search by name / category / endpoint description (case-insensitive) t2 services search # default limit: 10 t2 services search --limit 50 # broaden the result set t2 services search "" # list everything (empty query) # Inspect a single service or endpoint URL t2 services inspect # JSON output for scripting t2 services search "image" --json t2 services inspect --json ``` The CLI uses `T2000_GATEWAY_URL` (or `--gateway `) to override the gateway base URL — useful for local dev against `apps/gateway`. ## Example workflow ### "What AI chat models are available?" ```bash t2 services search "chat" ``` Returns a table of chat services (OpenAI, Anthropic, Gemini, Mistral, Cohere, DeepSeek, Groq, etc.) with cheapest endpoint price + base URL. ### "How much does GPT-4o cost?" ```bash t2 services inspect https://mpp.t2000.ai/openai ``` Returns every OpenAI endpoint with method + path + price + description. The user picks one (e.g. `/v1/chat/completions` at $0.01) and copies the URL into a `t2 pay ` call. ### "Send an email via Resend" ```bash t2 services search "email" t2 services inspect https://mpp.t2000.ai/resend ``` Lists email + messaging services; inspect Resend to see `/v1/emails` at $0.05. ## Output (default — search) ``` 3 services matching "chat": OpenAI from $0.01 [ai, chat] url https://mpp.t2000.ai/openai about OpenAI Chat Completions API Anthropic from $0.01 [ai, chat] url https://mpp.t2000.ai/anthropic about Claude messages API Mistral from $0.005 [ai, chat] url https://mpp.t2000.ai/mistral about Mistral chat completions Use `t2 services inspect ` to see pricing + endpoints for a service. ``` ## Output (default — inspect endpoint) ``` Service OpenAI URL https://mpp.t2000.ai/openai About OpenAI Chat Completions API Categories ai, chat Currency USDC on Sui POST /v1/chat/completions $0.01 Chat completions (gpt-4o, gpt-4o-mini) url https://mpp.t2000.ai/openai/v1/chat/completions Pay with: `t2 pay https://mpp.t2000.ai/openai/v1/chat/completions` ``` ## Output (--json) ```json { "query": "chat", "count": 3, "services": [ { "name": "OpenAI", "serviceUrl": "https://mpp.t2000.ai/openai", "description": "OpenAI Chat Completions API", "categories": ["ai", "chat"], "currency": "USDC", "chain": "Sui", "endpoints": [ { "method": "POST", "path": "/v1/chat/completions", "price": "0.01", "description": "Chat completions" } ] } ] } ``` ## When called through MCP (`t2000_services` tool) The MCP tool returns the full catalog JSON in one call (no search filter — the LLM filters in its head): ```json { "services": [ { "name": "OpenAI", "serviceUrl": "https://mpp.t2000.ai/openai", "endpoints": [...] }, { "name": "Anthropic", "serviceUrl": "https://mpp.t2000.ai/anthropic", "endpoints": [...] }, ... ] } ``` For LLM-driven flows, this is the right shape — the LLM scans the catalog, picks the matching service, and calls `t2000_pay ` next. ## Categories (live) The current catalog clusters into: | Category | Services | |---|---| | AI / chat | OpenAI, Anthropic, Gemini, Mistral, Cohere, DeepSeek, Groq, … | | AI / image gen | fal.ai, Stability AI, OpenAI DALL-E, Replicate | | AI / audio | OpenAI Whisper, ElevenLabs, OpenAI TTS | | Search | NewsAPI, Brave, Exa, Serper, SerpAPI, Jina | | Weather / maps | OpenWeather, Google Maps | | Finance | CoinGecko, AlphaVantage, ExchangeRate | | Translation | DeepL, Google Translate | | Code / utility | Judge0, screenshot-as-a-service, QR codes, PDFShift | | Email / mail | Resend, Lob (postcards, letters, verify) | | Commerce | Printful (print-on-demand), Hunter (email discovery) | | Security | VirusTotal | | Messaging | Pushover | | URL / IP | Short.io, IPinfo | > The categories above are a snapshot — the live source is `t2 services search ""` (lists everything). New services land regularly. ## Error handling | Error | Meaning | |---|---| | `GATEWAY_UNREACHABLE` | The gateway at `mpp.t2000.ai/api/services` is down or DNS is failing. Retry; if persistent, status page is at `t2000.ai`. | | `No services match` | The search query returned zero hits. Suggest a broader query or `t2 services search ""` to list everything. | | `No service matches ` (inspect) | The URL isn't in the catalog. Run `t2 services search` to find the right URL first. | ## What NOT to do - Don't hardcode service URLs in your prompts. The catalog is the source of truth. - Don't tell users a service costs "around $X" — call `t2 services inspect` and quote the exact price. - Don't bundle `t2 services search` + `t2 pay` into a single hidden step. Show the user what you found before paying.