Production proof
Products We Build and Operate
Running our own software in production keeps client work grounded in the parts that matter after a demo: reliability, billing, permissions, integrations, monitoring, and support.
ResearcherFlow
Literature in. Next experiment out. You approve. ResearcherFlow is the AI research workspace we operate in public — papers, questions, experiments, and writing in one place, with Flow proposing and a human deciding what is saved.
A real workspace, not a mock. Evidence stays beside the conversation. Flow waits for approval before writing into the project.
Who it is for
Graduate students, PIs, and scientists who currently keep literature, experiment logs, and writing in three different places.
What it does
Add papers by DOI, keep notes next to sources, plan the next test, and search the whole project. Export in Markdown, JSON, or CSV whenever you want it out.
What it costs
Free to start. Researcher is $29/month. Team is $99/month for a lab sharing workspaces. No sales call required to create an account.
What it takes to run in production
Nobody handed us a spec. We found the problem, scoped version one, and now run it. You can use it today. That is the most direct evidence we can give of the work we do for clients.
1,400+
Automated tests guarding the product
1,417
Commits shipped in the first five months
3 tiers
Live billing, permissions, APIs, and operations
- Django and Next.js on PostgreSQL, with Redis and Celery for background AI jobs
- Stripe billing across three tiers, with plan limits enforced in the product
- Workspace permissions, a public REST API, signed webhooks, and a Python SDK
- Sentry, PostHog, transactional email, and Docker-based deploys
What that means for your build
The specifics of your project will differ. The hard parts rarely do: billing that holds up when money moves, permissions that are enforced everywhere, AI grounded in your data, and operations that do not require someone watching a terminal.
If we can run a production SaaS with paid tiers, a public API, and a human approval gate, we can build the internal tool that is currently a spreadsheet.