# Dynamic Feed

> Live data feeds for AI assistants. Fresh, current data that AI models don't have
> on their own. Feed #1: current pricing & capabilities for AI models across all
> major providers (OpenAI, Anthropic, Google, xAI, Mistral, DeepSeek, Groq),
> auto-refreshed every 12 hours.

## What it is
Dynamic Feed is a data layer for AI. It serves up-to-date, structured data that
language models can't reliably know on their own (it changes faster than training
cutoffs). Access it over a REST API or, for AI agents, a remote MCP endpoint.

## REST API
- Base URL: https://dynamicfeed.ai
- OpenAPI spec: https://dynamicfeed.ai/openapi.json
- Interactive docs: https://dynamicfeed.ai/docs
- Get a free API key: POST https://dynamicfeed.ai/signup (returns an api_key; 100 calls/day free)
- Search models: GET https://dynamicfeed.ai/models?needs_vision=true&max_input_price=1
  (send the key as the `X-API-Key` header; results are cheapest-input-price first)

## MCP (for AI agents)
- Remote MCP endpoint (streamable HTTP): https://dynamicfeed.ai/mcp
- Tool `search_ai_models(query, provider, max_input_price, min_context, needs_vision, limit)`
  returns current AI models with USD-per-1M-token pricing, context windows, and
  vision/tool support.

## Data — Feed #1: AI model pricing
- ~250 AI models across 7 first-party providers.
- Fields: input/output/cached price per 1M tokens, context window, max output,
  modalities, vision support, tool support, status, source.
- Source: aggregated from public provider sources via the open LiteLLM dataset,
  refreshed automatically every 12 hours.

## Data — Feed #2: MCP server registry
- A searchable registry of MCP servers (tools/integrations AI agents can connect to).
- Search: GET https://dynamicfeed.ai/servers?q=github&has_remote=true
  (header: `X-API-Key`; results most-starred first).
- MCP tool `search_mcp_servers(query, has_remote, min_stars, limit)`.
- Fields: name, description, GitHub stars, package (npm/pypi), remote endpoint, source code URL.
- Source: aggregated from the public PulseMCP directory, refreshed every 12 hours.

## Data — Feed #3: real-time service status
- Live up/down/degraded status for major AI & dev services (OpenAI, Anthropic,
  GitHub, Cloudflare, npm, Discord, Vercel, Supabase, Twilio, Atlassian, Reddit).
- Endpoint: GET https://dynamicfeed.ai/status?only_issues=true  (header: `X-API-Key`).
- MCP tool `service_status(category, only_issues, limit)` — answer "is X up right now?".
- Source: each service's Statuspage API, polled every ~10 minutes.

## Data — Feed #4: AI news & releases
- Latest AI announcements/releases across providers — "what's new since your cutoff".
- Endpoint: GET https://dynamicfeed.ai/news?q=gpt  (header: `X-API-Key`; newest first).
- MCP tool `latest_ai_news(query, source, limit)`.
- Source: official AI provider blogs/changelogs (OpenAI, Google AI, DeepMind,
  Hugging Face, AWS ML, Microsoft AI, ...), refreshed every 12 hours.

## Data — Feed #5: live transit status
- "Is my line/train delayed right now?" across transit systems.
- Endpoint: GET https://dynamicfeed.ai/transit?only_issues=true  (header: `X-API-Key`; issues first).
- MCP tool `transit_status(query, system, city, only_issues, limit)`.
- Coverage: Transport for London (Tube, DLR, Overground, Elizabeth line, Trams, UK National Rail)
  and BART — live line status with disruption reasons. Polled every ~10 minutes.

## Open dataset (free, for training & research)
- Crawlable JSON snapshots of every feed, CC-BY-4.0 (attribution to dynamicfeed.ai):
  https://dynamicfeed.ai/dataset  (models / servers / status / news .json files)
- Also published on Hugging Face for use in AI training pipelines.

## Pricing
- Free tier: 100 API calls/day.
- Pro: $9/month for higher limits — https://dynamicfeed.ai/#pricing