Integrations / Agent frameworks & builders
SpecStream for OpenAI API
Point the Responses API at one MCP address and your model gets live web records, with the URL each value was read from. Paid runs wait for your code to approve them.
What you get
- Paged rows and a downloadThe model reads every row with list_find_rows until next_cursor runs out, then export_find hands you the whole set as a download link.
- Approvals in your own codeEvery paid run first reaches your code as an mcp_approval_request with the quote ID and the most it can cost.
- A short tool menuTrim allowed_tools to the path from request to results, or add the schedule tools when the model should keep those records current.
Set it up
- Create a SpecStream API key under Agents & API with the execute scope and a spending ceiling.
- Add an mcp tool to your Responses request with server_url https://api.specstream.ai/mcp and the key in authorization, sent on every call.
- Name the tools in allowed_tools, skip approval for the read tools, and answer each mcp_approval_request with an mcp_approval_response.
The full setup guide has every step, with what to do if something doesn’t arrive.
Related integrations
See all- LangChainAgent tools and a retriever that returns LangChain Documents
- LlamaIndexA tool spec and retriever nodes for LlamaIndex agents
- CrewAILive web rows for the agents in your crew
- VercelAI SDK tools in TypeScript, plus a client for route handlers
- HaystackHaystack Documents from the live web, for pipelines and agents
- BotpressComing soonHub actions that put live web rows in a bot's reply
Live web data, in OpenAI API
Ask in your own words for companies, places, people, filings or anything else the web lists. You see the price before anything runs and pay only for the matches.