About flintspark
flintspark is an enterprise AI platform. Flint, flintspark's agent, learns a business workflow by watching a person do it once in the browser, checks it with verified runs, then runs it in web tools. By default a person approves every irreversible action.
- Name
- flintspark
- Website
- flintspark.ai
- Category
- Enterprise AI platform
- Agent
- Flint, flintspark's agent
- Status
- In development
- Deployment today
- One operator's machine
- Certifications
- None
- Location
- San Francisco Bay Area
What does flintspark do?
It turns browser work into governed AI agents. A person shows Flint a task once; Flint records the steps, maps the screens and learns a skill.
A new skill is trusted only after two verified runs. It then runs on the fastest route that works, with or without an API, and keeps each run's steps and evidence for review.
From one demonstration to a trusted skill
- ShowA person does the task once in a browser.
- VerifyThe skill must pass two verified runs.
- RunFlint runs it; by default a person approves what cannot be undone.
What do we believe?
flintspark is named for what flint does: it doesn't burn by itself. It strikes, and the customer's work catches fire. We are the spark; the customer is the fire and always the hero.
- People approve. By default, a person approves every irreversible action, and approvals go into a tamper-evident audit log.
- Policy decides. What Flint may do is set by policy, never by the model, and learning a new skill cannot loosen it.
- Evidence over promises. Every run keeps its steps and evidence, and we label what is built, in build or on the roadmap.
Two rules that do not change
People approve.
Policy decides.
Release, disposition and signing are never run by Flint, in any mode.
Where does flintspark stand today?
flintspark is in development. Current demos cover a clinical-trial and FDA watch and energy procurement; support for pharma manufacturing quality review is in build (roadmap). flintspark is not a validated GxP system and does not release product on its own.
Today it runs on one operator's machine. flintspark holds no certifications, and none is committed yet.
How we check what we publish about it: our editorial and AI-use policy.
Status today
- Stage
- In development
- Runs on
- One operator's machine
- Certifications
- None
- Pharma lot release
- in build
Questions about this page
Is flintspark available to buy today?
No. flintspark is in development. It runs today on one operator's machine, and this site shows demos of Flint, flintspark's agent, on example workflows. It lists no pricing, customers or team.
What does flintspark believe about AI agents?
That people should approve what cannot be undone, and that policy, not the model, should decide what an agent may do. By default a person approves every irreversible action, and approvals go into a tamper-evident audit log.
Is flintspark related to Apache Spark?
No. flintspark is enterprise AI software at flintspark.ai. Its agent is Flint, flintspark's agent. It is not related to Apache Spark or to other products named Flint.
See Flint run
Bring one workflow you'd rather not do by hand. We'll show Flint learning one like it, and stopping before anything it can't undo.
We reply from hello@flintspark.ai.