Live data tools for Pydantic AI agents.
Point your Pydantic AI agent at one MCP URL and it gets 87 keyless tools of fresh data the model was never trained on - weather, CVEs, service status, LLM pricing, rates, hazards - each answer signed and stamped with its source.
The call
Test the keyless endpoint your agent's tools will hit - same request a Pydantic AI tool call makes, no key:
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
Pydantic AI agents are typed and tool-driven, but a tool is only as good as the data behind it - and an LLM's weights are frozen at training time, so it cannot know today's CVEs, current software versions, live service status, or this morning's rates. Dynamic Feed is a keyless MCP server (Streamable HTTP) plus a /v1/batch REST endpoint that hands your agent 87 tools across 19 verticals drawn from 70 sources. Connect via the remote MCP URL and the tools register automatically; or call /v1/batch directly from a Pydantic AI tool function when you want explicit, typed control over which feeds run. Every datapoint comes back Ed25519-signed and stamped with provenance: source, observed-at, and freshness. Be precise about what that signature means: it is proof-of-existence and tamper-evidence - proof that Dynamic Feed reported value X 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 owns judgment; the signature just makes the input auditable. Note there is no live crypto, metal, or stock price conversion here - that is paused. The finance tools (interest_rates, treasury_yields, global_macro, china_data) are live macro and rates data, not market prices.
Use it for
- Register 87 live-data tools in a Pydantic AI agent with one MCP URL, no key
- Call /v1/batch from a typed Pydantic AI tool function for explicit feed control
- Ground agent answers in citeable, signed facts so outputs are auditable
- Catch model staleness at runtime with reality_check and software_version
FAQ
Do I need an API key?
No. The MCP endpoint at https://dynamicfeed.ai/mcp and the POST https://dynamicfeed.ai/v1/batch endpoint are both keyless and free for fair use. Connect and call - no signup.
How does this fit Pydantic AI's tool model?
Two ways. Point an MCP client at the remote URL and all 87 tools register automatically, or wrap a /v1/batch call inside a typed Pydantic AI tool function when you want to validate args and control exactly which feeds run.
What does the Ed25519 signature actually guarantee?
It is proof-of-existence and tamper-evidence: proof we reported a value at a given time and that the response was not altered. It is not a claim the value is true, accurate, or safe - your agent keeps that judgment.