Agent-Native Financial Analysis with Orcaset

This video below introduces Orcaset through a simple office acquisition model. Orcaset is designed for agents to build open financial models that leverage data from outside systems and scale to portfolio-level analysis.

The introduction highlights key Orcaset features that traditional spreadsheet can’t reproduce.

Reusable Components

Models are built up from logical components. Components can be bundled into external, well-tested modules that are reused across models. Orcaset’s type safety ensures components fit together properly, raising an error if the model tries to combine incompatible values.

The video models a lease-up scenario, reusing the existing Lease component for the new tenant.

Inline Data Retrieval with Citations

Users can define values with arbitrary functions, including functions that fetch data from external resources such as vendor APIs, internal analytics apps, or databases. Resource connections can be managed securely using software best practices for managing authorization tokens.

Values can also take any shape. For example, numbers can carry metadata with their source citations.

The video defines a market-rent function for the new lease that searches the internet for an applicable broker report and parses current office asking rents.

Version Control

Orcaset models work natively with standard code version-control software like Git and GitHub. Users can easily fork model scenarios, switch between branches, and recombine them. Each saved checkpoint is stamped with the author’s name, creating an audit trail of who made which changes over time.

Explore Orcaset

Get started with Orcaset by sending the prompt below to Claude or ChatGPT, or examine the base files from the video here.