Nothing enters EkamFlow by upload. Data is read from the connection you gave it, for the job that needs it.
Every answer is reachable by API.
The console is one way in. A profile can be pulled by your systems, a decision asked for while a customer is on the page, a workflow driven from a notebook, and EkamFlow read from the assistant your team already uses — each under a scope that cannot reach past itself.
Six ways in, each scoped.
A key carries a scope and is refused on any route outside it, and is never treated as an admin when database and column access is decided.
Profile API
Fetch a customer's current row, or the whole profile after each refresh, from your own systems.
GET a profile · GET one entity's row · POST a download, then read its files
- One customer's row by key, or the whole profile as CSV or Parquet files.
- An API key scoped to profiles reaches a profile's own routes and nothing else.
- A key sees what a person with read access to that profile sees: hidden and masked columns stay that way.
- Each answer reports the profile version and when it was last refreshed, so you know how fresh your copy is.
- The profile's API tab test-runs both calls and shows the curl and Python for them.
Decisions API
Ask what to do for the customer in front of you, at a moment you defined.
POST a decide call for one customer at one decision point
- Your server asks for a customer at a named decision point and gets back one of the actions you declared, or no action.
- An API key scoped to decisions cannot read or write anything else.
- A share you choose is held back from the decision so the effect can be measured.
- Cooldowns and a contact cap are enforced on EkamFlow's side, not left to the caller.
- Every decision is logged, and the log clears if you withdraw approval to store context.
Python SDK
Everything the console does, from a notebook or a scheduled script.
- Define schemas and use cases, start training and prediction runs, and read results.
- Query the databases a run wrote, and save queries for later.
- List, invite and manage members with a person's login — an API key cannot manage people.
- The SDK and the console stay in step: the same operations, checked against each other.
EkamFlow from your own assistant
Use EkamFlow from Claude, ChatGPT or Cursor, as yourself.
- Sign in once from your assistant and ask it to list workflows, read predictions and segments, check health, or query databases.
- It sees exactly what you see in the console, and nothing more.
- Anything that spends money or writes to your systems still waits for someone to press Start in EkamFlow.
- Off by default until an admin turns it on. The admin then sees which assistants are connected and can revoke any of them.
Your tools inside AI Studio
Connect Snowflake, Databricks, HubSpot, Salesforce, BigQuery, GitHub, Linear or your own server.
- Connect with your own access token, or by signing in where the service supports it.
- Each tool call waits for your approval before anything is sent.
- Admins can switch member connectors off, or allow only listed hosts.
- A connector can be switched off for a single chat without disconnecting it.
Keys and access
Scoped keys, revoked in one place, audited throughout.
- A key carries a scope — read, predict, profiles, decisions, admin — and is refused on any route outside it.
- A profiles key reaches a profile's own routes and nothing else; a decisions key reaches one route.
- Whatever its scope, a key is never treated as an admin when database and column access is decided.
- Revoking a key stops work already queued under it, not just the next call.
- Access is checked per role, per database and per column tag, and masks apply before any query runs.
Your warehouse stays the system of record.
Predictions are written back through your own warehouse and your own tools, so your systems stay the system of record.
Build an Agentic Worker for your team.
Tell us which workflow you want to hand off.