MiniApps
What MiniApps are
MiniApps turn financial research workflows into focused hosted tools without requiring a separate engineering project.
The category
MiniApps are small hosted tools you build by describing them to Kai: a hypothesis intake, evidence library, backtest review dashboard, risk assessment, investment memo workflow, or research tracker.
Code is the substrate. You never see it. You describe what you need. Kai builds and deploys it on your workspace, on your domain, wired to your workspace data, email, files, and clients.
- ›Build a hypothesis intake with universe, signal, costs, benchmark, and rejection rule.
- ›Create an evidence library with citations, paper roles, and visible limitations.
- ›Build a portfolio review dashboard with exposures, benchmark context, and as-of timestamps.
Why this is different from page builders
Page builders make pages. MiniApps run workflows.
A static page presents information. A MiniApp runs a workflow: researchers submit hypotheses, reviewers record decisions, analysts compare evidence, and teams inspect live study state.
One workspace, one source of truth.
Your studies, papers, datasets, documents, and MiniApps live in the same workspace. A research intake can write directly to its namespaced data store and link back to workspace evidence.
You iterate by talking.
"Add transaction-cost assumptions to the review form." Kai changes the app and redeploys it. The workflow can evolve alongside the research process.
What you can build
The pattern that repeats across financial teams: a focused workflow that is too specific for a generic form or dashboard.
- Quant researcher. Hypothesis intake, evidence libraries, study pipelines, backtest review dashboards.
- Portfolio manager. Mandate intake, exposure dashboards, risk assessments, committee review flows.
- Risk analyst. Control assessments, exception trackers, model-documentation portals.
- Investment analyst. Valuation calculators, diligence portals, investment memo workflows.
See Use cases by persona for full examples.
How a MiniApp gets built
The end-to-end flow, in plain language:
- 1. You describe what you want. "Build a review dashboard for completed backtests."
- 2. Kai uses your research context. Mandate, methodology, datasets, studies, papers, and review state.
- 3. Kai picks a starting template. One of the prebuilt MiniApp templates. See Templates.
- 4. Kai customizes the code in a sandbox. You watch the changes pile up as a draft PR.
- 5. You preview it. The app runs in your runtime under a preview URL.
- 6. You ask for changes in plain language. Kai edits and redeploys.
- 7. You publish on your domain. Tell Kai the hostname (e.g.
research.example.com). If the domain isn't connected yet, Kai walks you through DNS and HTTPS. See Add a domain. - 8. You iterate. "Add a benchmark comparison and limitation field."
What every MiniApp can do
MiniApps aren't sandboxed pages. They're first-class citizens of your workspace. Via the MiniApp SDK, your app can:
- Read and write its own data (namespaced Mongo collection per app).
- See who the caller is (your analyst, reviewer, or client, with auth-checked identity).
- Send email from your domain.
- Store and serve files (private or public).
- Hold approved per-company integration secrets without exposing them in code.
- Call Kai from inside the app to summarize evidence, draft a review note, or explain a submission.
The contract is small and stable. See MiniApp SDK.
The promise
MiniApps are for interactive workflows and client operations. They are not static websites or payment surfaces.