Products
Temper Labs is developing two products: Finstruct, an agent harness, and Forge, our evaluation platform.
finstruct
Finstruct is a customizable, open-source agent harness for capital markets work. It integrates easily into our Forge evaluation platform.
- Provider Neutral
- Finstruct can be configured to work with any model provider.
- Easily Extendable
- Anyone can add new tools, large language models, mcp servers ro their Finstruct harness. Finstruct integrates skills to allow users to orchestrate their capital markets work.
- Fully Traceable
- Finstruct is designed for full observability so that financial institutions can audit and understand what actions models are taking.
forge
Forge is the Temper Labs evaluation platform. It turns capital markets workflows into clear datasets so institutions can assess model intelligence, compare harnesses, and run reliable evaluations.
- Dataset Management
- Define tasks, metadata, and rubrics so evaluation work is structured, versioned, and ready for repeatable runs.
- Evaluation Runner
- Execute evaluations across models and harnesses with consistent scoring against real capital markets jobs to be done.
- Analytics and Insights
- Compare results across providers and configurations to see which models are good for which financial tasks.