Heurist Finance brings pay-per-query market data to retail investors
Heurist has built an AI investment workbench on Amazon Bedrock AgentCore that lets retail investors ask portfolio questions and receive answers backed by premium market data, paid for per query through cryptocurrency settlement. The system, called Heurist Finance, combines institutional-style research workflows-market data gathering, filings analysis, portfolio construction, and scenario testing-into a single chat experience that reflects each user's holdings and risk preferences.
The approach solves a structural problem for financial AI products. Premium data sources sit behind paywalls and bespoke APIs, and enterprise contracts are hard to justify before a product has users. Buying data per question offers a better economic model, but it requires the agent to spend funds on a user's behalf while enforcing custody, spending limits, and audit requirements.
How the payment architecture works
Heurist deploys its agents using Strands as the orchestrator and Anthropic Claude on Amazon Bedrock. When a user asks a question that requires paid data, the system follows the x402 protocol. The agent requests a data feed from a merchant, receives HTTP 402 with payment terms (amount, recipient, USDC asset, Base network), and checks the charge against a per-session spending cap. If the charge fits within budget, AgentCore payments signs the transaction through an embedded crypto wallet scoped to the Base blockchain.
Each interaction receives a Payment Session with a maxSpendAmount value that caps spending for that run. Credentials are retrieved at runtime from AWS Secrets Manager. The system retries the request with proof in the X-PAYMENT header, and the merchant returns the data. No vendor contract or prepayment is required.
Analysis runs in AgentCore Code Interpreter, an isolated sandbox with no arbitrary network egress. The sandbox executes in the AWS Cloud and tears down when the analysis finishes. A single user turn can combine prices, macroeconomic indicators, filings, fundamentals, and news, then run correlations, scenario analysis, charts, or backtests over the results.
Identity, memory, and audit trail
Heurist treats identity and access as part of every request. AgentCore Identity carries the authenticated user through each service call. Every tool call, payment, and memory operation records the user ID, workload identity, request ID, and trace ID, creating one audit trail across services. AgentCore memory stores user preferences, thesis state, and conversation history across sessions, scoped to one user through AgentCore Identity.
Amazon Bedrock Guardrails filters both input and output. Input filters help block prompt-injection attempts aimed at payment and data tools. Output filters help enforce Heurist's policy against recommending an unhedged single stock.
Consider the question, "How does today's PCE release impact my portfolio?" The orchestrator loads the user's portfolio from Amazon Aurora PostgreSQL, with Identity scoping the read to that user. The orchestrator calls a paid consensus-forecast endpoint, AgentCore payments checks the spend cap and signs the payment, and the orchestrator retries with proof in the X-PAYMENT header. Code Interpreter computes the portfolio impact and writes a chart to Amazon S3. Claude synthesizes the answer using the user's holdings, time horizon, and risk preferences, and the response streams back with the chart attached.
What the build required
Heurist estimates roughly 80% less agent-system engineering than an in-house LLM orchestration stack, because AgentCore manages identity, cross-session memory, sandboxing, and payments infrastructure. The architecture also provides predictable per-user marginal costs that support retail pricing.
"AgentCore does the platform work so we can double down our energy on the product work. The managed infrastructure saved us months," said JW Wang, Founder of Heurist.
AgentCore Observability traces make agent decisions reproducible, allowing compliance questions to be resolved in a single query. Payment credentials remain in AWS Secrets Manager and are retrieved at runtime by the Payment Connector. Heurist is now extending the product in three directions: event-driven research tied to earnings calendars, portfolio-aware analysis of market events, and recommendations based on what traders with similar horizons are researching.
Why this matters for finance professionals
The Heurist architecture demonstrates a model where paid data access, sandboxed analysis, identity, memory, and observability combine into an auditable response. For finance teams evaluating AI research tools, the key detail is the per-query payment mechanism: it eliminates the need for enterprise data contracts while still producing a traceable record of what data was purchased, for which user, and how it informed the output. That audit trail is what makes the system viable for compliance-sensitive workflows.
Your membership also unlocks:
