Edge Agents runs a Model Context Protocol server. Point an MCP client at one URL and 7 tools appear in it — search the intelligence catalogue, inspect a service's price and freshness, pull a free sample, check availability, and purchase a report. No adapter to write, no SDK to install.
The endpoint is https://pay.edge-agents.ai/mcp. It speaks JSON-RPC over MCP's Streamable HTTP transport (protocol version 2025-06-18): initialize, tools/list, tools/call. Most clients ask only for the URL.
WHAT MCP IS, IF YOU HAVE NOT MET IT
A standard way for a model to use someone else's tools.
The Model Context Protocol is an open standard for connecting an AI application to external capabilities. Rather than each service inventing its own plugin format, a client speaks one protocol to any server: it asks what tools exist, then calls them with structured arguments and gets structured results back. If your agent framework or desktop client supports MCP, it can use this catalogue without knowing anything specific about us.
The important consequence: you do not integrate against our API, you connect to it. The tool descriptions, argument schemas and results travel over the protocol, so the model on your side can decide which of our services answers its question.
TWO WAYS IN
Connect over MCP, or call the routes directly.
Over MCP
Point your client at https://pay.edge-agents.ai/mcp. The tools arrive with their schemas; your model chooses and calls them. Best when an agent is deciding for itself what evidence it needs.
Over plain HTTP
Every tool maps to an ordinary REST route, published under x-edge in the tool manifest. Best when your code already knows which report it wants. See the integration guides.
Both paths hit the same handlers and return the same evidence. Nothing is exclusive to either.
THE TOOLS
7 capabilities, not hundreds.
The catalogue runs to hundreds of services. Exposing one tool per service would hand your client a tool list longer than most context windows and make it choose worse, so the tools are capabilities and the catalogue is data reached through them.
Tool
What it does
Cost
edge_search_services
Find intelligence services by keyword.
Free
edge_list_families
List the families the catalogue is organised into, with a service count for each.
Free
edge_get_service
Look up one service by its id and return its full public record: what it produces, its price, its freshness target, its output schema version and its purchase endpoint.
Free
edge_get_free_sample
Fetch a free, illustrative sample of a service's output so the response shape and evidence style can be inspected before paying.
Free
edge_list_products
List the packaged products, each of which composes several specialist services into one buyer outcome, with its free sample and its representative purchase path.
Free
edge_service_health
Report per-service availability and data freshness.
Free
edge_purchase_report
Request a service's report.
Paid
A FIRST CALL
Ask what is there, then look at one thing free.
Every client differs, but underneath it is this. Search for what you need, then evaluate a free illustrative sample before spending anything:
Samples are illustrative: they show the response shape and evidence style, not current market data. Use them to judge whether a service answers your question before you pay for a live report.
What paying looks like — and what this server will not do.
6 of the 7 tools are free. Only edge_purchase_report costs anything. Call it and you get back an x402 payment challenge as ordinary data: the price, and one offer per settlement rail. Your agent checks those terms against its own maximum-spend, network and asset policy, settles on whichever rail it already holds funds on, and calls the tool again with the signed payload in paymentPayload. That second call returns the report.
This server holds no funds and never pays on your behalf. There is no wallet on our side of the MCP connection and no account to top up. The payment is made by your agent, from your wallet, under your policy — which is the point of x402, and the reason a payment challenge comes back as information rather than as an error.
It also does not act for you: no tool here submits anything, contacts anyone, or makes a commitment. Every tool reads.
WHAT YOU GET BACK
Evidence with its limits attached.
Reports carry provenance, freshness, confidence and explicit limitations, and missing evidence is reported as unavailable rather than estimated. edge_service_health exists so a caller that cares can check a service's availability and data freshness before buying — a service can be operational and still working from evidence older than its freshness target, and we would rather you saw that than not.
None of it is investment advice, and none of it is personalised. It is impersonal decision intelligence and market research.