AI Exchange Node is a standard, secure MCP connector, IN and OUT, that joins your organisation to a trusted exchange. Your agents can work with partners' systems, and partners' agents with yours, under rules everyone shares.
How the exchange worksAn exchange is a group of organisations whose nodes all work the same way and who trust one common trust anchor to vouch for every member. Join once, and you can work with every member.
Every pair of organisations builds its own integration, swaps its own keys and signs its own contract. Six organisations already need fifteen of them.
Each organisation runs one node and is vouched for once by the trust anchor. Any member can then call any other directly; the anchor confirms who is who but never sees the data.
Every node speaks MCP and follows the same trust framework, so any member can connect to any other without custom work.
An association, regulator or consortium vouches for who the members are, and can withdraw that at any time.
Information goes straight from one organisation to another. Each keeps its own data, its own rules and its own evidence.
The node is the organisation's single, standard gateway into the exchange, for traffic in both directions.
Three standard endpoints, from open to restricted: public (T0) for anyone, customer (T1) behind an OAuth login, and partner (T2) for agents that present a credential and a mandate. Each endpoint shows only the tools you attach to it; adapters connect those tools to your systems.
Your AI agents live on the node, each with its own identity and mandate, and call partners' nodes through it. The outbound trust gateway checks each partner's membership, revocation status and key before any call leaves.
The Model Context Protocol connects AI agents to tools, but it does not say who is on the other side. MCPF (MCP Trust Framework) adds that: verifiable identities for organisations and agents, credentials from the trust anchor, and revocation that reaches everyone.
Every AI Exchange Node implements MCPF, which is what lets organisations that have never met trust each other's agents from the first call.
Every organisation and every agent has its own verifiable identity, tied to the organisation responsible for it.
The trust anchor issues each member a signed credential. Any node can check it without asking anyone.
A node proves on the spot that it holds its own key, so published documents alone cannot be used to impersonate it.
Each agent carries a mandate from its organisation. Withdraw the mandate or the membership and access ends everywhere, even mid-session.
It shows who it is, which organisation it belongs to and what it is allowed to do.
Identity, membership with the trust anchor and the mandate. If anything fails, the call stops.
Only what the mandate allows, and both sides keep a signed record of what happened.
Registers, insurers, utilities, public bodies. Share selected functions with partners' agents without building an API programme.
Give each agent one identity and one mandate, and reach every member without separate contracts or keys.
ChatGPT, Claude and similar assistants sign in with the person's normal company login and stay within the rights the company gave.
Associations, regulators and consortia set the rules, admit members, and withdraw access across the whole exchange at once.
The node does not decide whether an agent is good. It makes an organisation's word about its agent — who it is, what it may do, until when — checkable by every partner and revocable everywhere at once.