Live data for the OpenAI Agents SDK.
Your OpenAI Agents SDK agent only knows what was in its training cut-off. Point its MCP integration at one keyless URL and it gets 87 tools of fresh, current data, each datapoint signed and stamped with source and observed-at time.
The call
No key needed. Try the same tools your agent will call, straight from a terminal. Each result comes back with an Ed25519 signature and a provenance block (source, observed-at, freshness):
curl -s -X POST https://dynamicfeed.ai/v1/batch \
-H 'Content-Type: application/json' \
-d '{"calls":[
{"tool":"reality_check","args":{"claim":"the latest stable Python is 3.12"}},
{"tool":"software_version","args":{"product":"python"}}
]}'Sample response
{ "mcpServers": { "dynamic-feed": { "url": "https://dynamicfeed.ai/mcp" } } }Why live data
The OpenAI Agents SDK can attach MCP servers as tools the model calls at runtime. The problem is the model's knowledge stops at its training cut-off, so it guesses at anything current - today's CVEs, whether a service is down, the latest package version, live macro and rates figures. Dynamic Feed is the keyless MCP server that fills that gap: register one URL and the agent gets 87 live tools across 19 verticals drawn from 70 sources. Every datapoint is Ed25519-signed and provenance-stamped with its source, observed-at time, and freshness. Be precise about what the signature means: it is proof-of-existence and tamper-evidence - cryptographic proof that we reported value X from source S at time T, and that the bytes were not altered in transit. It is not a claim that X is true, accurate, or safe. Your agent still decides how to use the fact; we just make the fact verifiable and traceable to its origin. Note there is no live crypto, metal, or stock price conversion here - the finance tools are macro and rates (interest_rates, treasury_yields, global_macro, china_data), not market prices.
Use it for
- Stop an Agents SDK agent answering current questions from a stale training cut-off
- Ground a tool call in a citeable, signed fact with a source URL and timestamp
- Give an agent live awareness of CVEs, service outages, software versions and weather hazards
- Hand the agent verifiable data it can pass through to an audit log or downstream check
FAQ
How do I add this to the OpenAI Agents SDK?
Register Dynamic Feed as an MCP server. The endpoint is https://dynamicfeed.ai/mcp over Streamable HTTP, keyless. Use the SDK's hosted MCP tool or a Streamable HTTP MCP server pointed at that URL, and the 87 tools appear to your agent automatically. Prefer plain HTTP? POST a batch of tool calls to https://dynamicfeed.ai/v1/batch, also keyless.
Do I need an API key?
No. Both the MCP endpoint and POST /v1/batch are keyless and free for fair use. There is nothing to sign up for to try it.
What does the Ed25519 signature actually prove?
It proves we reported a given value from a named source at a given time and that the bytes were not tampered with. It is proof-of-existence and tamper-evidence, not a claim that the value is true, accurate, or safe. Your agent still decides how to act on the fact.