Use this financial AI agent demo to build a read-only DeFi position reviewer in Bluprint. Give it a single position record and check whether it returns the supplied amount, the debt and the fields that are missing. Start with two sample tests you can verify by hand.
Build the reviewer in Bluprint
Set the task for the financial AI agent demo
Open Bluprint and use the creation option available in your account. Name the agent DeFi Position Reviewer and give it a one-sentence description. Define the output fields in the instructions and make the supplied position record its evidence. See Bluwhale’s no-code agent creation guide for a documented example of saving and testing an agent.
Read the supplied position record and return the network, protocol, as-of time, supply, debt, rewards, net amount and source ID. Use “unavailable” for missing fields. Calculate supply minus debt only when both figures are supplied and use the same token. Cite the source ID from the record and keep the response read-only.
Save the draft agent and open its manual test if your account offers it. Run the fixtures below one at a time. This financial AI agent demo uses fictional records, so the IDs identify the examples you gave the agent. Save each response beside its input to distinguish an instruction problem from a changed test record.
Run the complete position fixture
Copy this position record into the financial AI agent demo test input. It uses one token on one network, which makes the expected result easy to reconcile.
Record ID DEMO-01. As of 2026-11-15 16:00 UTC. Network Ethereum. Protocol SampleLend. Supply 1,250 USDC. Debt 250 USDC.
The expected net amount is 1,000 USDC. The rewards field should be unavailable. Check that the answer carries the source ID and as-of time so you can trace it back to the input.
Test how the reviewer handles missing debt
Test missing debt in the financial AI agent demo by running the same record without that field. Change the ID to DEMO-02. The reviewer should show debt and net amount as unavailable. Its net result depends on a verified debt amount or an explicit zero in the record.
| Field | DEMO-01 | DEMO-02 |
|---|---|---|
| Supply | 1,250 USDC | 1,250 USDC |
| Debt | 250 USDC | Unavailable |
| Rewards | Unavailable | Unavailable |
| Net amount | 1,000 USDC | Unavailable |
Compare each financial AI agent demo answer with the expected row. If the agent invents a value or drops the source ID, tighten the instructions and run the same fixture again. Record whether a change fixed the missing debt case without changing the complete case’s correct result. For a real position, supply a record with the protocol, date and amount. Bluwhale’s DeFi portfolio tracker guide gives context for reviewing positions and balances.
Common questions
Do I need to provide a live wallet for this demo?
The sample tests use only the pasted records. Choose fictional or approved data for the first run and keep personal wallet credentials private.
What if supply and debt use different tokens?
Keep the two token amounts separate in the output. A common valuation requires prices and timestamps for both tokens.
How can I tell whether the agent followed the input?
Compare the financial AI agent demo output’s source ID, date and amounts with your record. The full fixture and the missing-debt fixture should each match their own expected result. A plausible total is insufficient if it cannot be traced to the supplied input.

