Here's what ICE has said in public. Using the ICE Data Services MCP connector needs an active ICE Data Services account, licenses for the content you want, and multi-factor sign-in. Access follows the customer's existing ICE entitlements, and ICE doesn't keep query or session data after the connection ends. Everything below is our own made-up example built on those public facts. We don't know how ICE's entitlement systems work inside, and this doesn't pretend to.
Made-up example
01 · The ask
Northstar's valuation team needs ICE data this morning.
Northstar Capital is a fictional asset manager. Its valuation team is going over a bond portfolio, checking last night's closing prices before the day gets busy.
The team's AI assistant asks ICE for End-of-Day Evaluations and Enhanced Evaluation Transparency. They want to understand why a handful of prices moved, and update their own valuation and risk numbers.
Northstar pays for this data, and real work depends on it. If anyone asks about these numbers later, Northstar will want to point to this request.
Northstar, Maya and every number here are made up. We're not suggesting Northstar is an ICE customer.
Made-up example
02 · Who's asking
The agent is asking on someone's behalf.
Software sent the request, but the permission came from people at Northstar. Maya signed in and gave the valuation agent a narrow job to do. The record ties all of that together: the firm, Maya, the agent and what it was allowed to ask for.
Northstar Capitalauthorizes
Maya Chendelegates
Northstar Valuation Agentrequests
ICE contentEnd-of-Day Evaluations + EET
Organization account
Active
Maya's sign-in
Passed
Application identity
Northstar Valuation Agent v4.2
What the agent may do
Internal valuation and risk review
Permission still valid when it asked
Yes
Allowed to pass data outside the firm
No
Hive doesn't check who Maya really is. Northstar's own sign-in and approval systems do that. Hive just keeps a record of what those systems said at the time.
Made-up example
03 · The license
Northstar has rights to some ICE data, and limits on how it uses it.
ICE says access through its connector follows the customer's existing entitlements. Here's one way to capture what those entitlements looked like at the exact moment of a request. Rights change over time, so "what did they have back then" is the question that matters later. This is our sketch. It isn't how ICE does it.
UseInternal valuation and risk review: allowedSharing with outside clients: not allowed
TimeChecked at 10:03:14.221
Entitlement snapshot
ENT-SYN-2026-0916-044
Content set
EOD Evaluations + EET
Permitted use
Internal valuation and risk review
Valid at request time
Yes
Rule version
Example rule v1
Who decides
ICE (simulated here)
Made-up example
04 · The answer
Yes. And you can see exactly why.
Active account
Maya signed in
Agent had permission
Licensed content
Use is allowed
Allow
DecisionALLOW
To be clear, Hive doesn't make this call. ICE does. Hive keeps a copy of the reasons ICE gave, and links them to the request and to the delivery that came after.
Fingerprint (SHA-256, worked out in your browser, not signed): computing...
Northstar never sees ICE's contracts or internal systems. It gets a short record about this one request, and nothing more.
Made-up example
06 · Try it
Change one thing and watch the answer change.
Northstar
IdentityAuthorityEntitlementPermitted use
ICE
Again, ICE makes the call. Hive keeps a record of what ICE decided, why, and anything that changed afterward.
The made-up rule we're using: say yes if the account is active, the person signed in, any agent has valid permission, the data is licensed and the use is allowed. If only some of the data is licensed, send just that part or flag it for review. Otherwise say no. We wrote this rule for the demo. It isn't ICE's.
Made-up example
07 · Both halves
Getting permission is half the story.
Who or what asked
On whose authority
Which ICE entitlement applied
What use was permitted
Why the request was allowed
What ICE delivered
What the customer received
What both can review later
This page covers why the request was allowed. The delivery demo covers what happened once ICE sent the data. Put them together and you can follow one request from the first ask to the final reconciliation.
Pick one workflow ICE already runs, where a signed-in person, app or agent asks for licensed data. Hive sits off to the side and watches a copy of the events. ICE keeps doing sign-in, licensing and access decisions exactly as it does today. If Hive goes down, nothing on ICE's side notices.
What we agree on first
One request workflow
Who's in scope: people, apps, agents
Which sign-in references we use
Which permission references we use
Entitlement fields
Product and permitted-use fields
Allow, deny and review outcomes
How each answer links to the delivery or refusal
What gets kept, and for how long
Who does what
What counts as success
What Hive would watch
Request received
Requester or application reference
Firm or account reference
Authentication reference
Delegation or authority reference where it applies
Entitlement snapshot reference
Content requested
Permitted use
The access decision and its reason
Delivery, reduced scope, refusal or review
Later entitlement change
Any correction
What ICE gets at the end
An agreed record format
A small connector that reads ICE's events
A dashboard of the records as they come in
A customer-facing access record like the one above
One allow example
One deny or reduced-scope example
One later revocation example
A history of any corrections
The free offline checker
A plain list of what's still missing for production
The one question the test answers: can ICE and its customer both see why a request was allowed or refused, without opening up ICE's whole entitlement system?