Two products. One boundary.

Promptev is privacy-first, context-aware AI for teams. The App is where your people build agents and talk to them. The Context engine is the retrieval library underneath — the part that decides what an AI is allowed to know, and it runs perfectly well on its own.

Start free Talk to us

2,500 free credits · no card · model calls go to your provider on your own keys

Ready to connect: 16 servicesEight apps to approve · 198 tools
Beyond those: Anything with an APIHTTP tools, database tools, your own MCP servers
Where it runs: Your callManaged by us, your cloud, or on-premises
The engine: Apache-2.0Python and TypeScript, one PostgreSQL schema

The product your team opens

Promptev App

Where a team builds and runs agents. A workspace holds your projects; a project holds its knowledge, its tools and the agents that use them. People ask in the browser, in Slack, in WhatsApp or in a widget on your own site, and every answer names the documents it came from.

Knowledge with per-caller permissions. Everyone asks the same agent; nobody gets an answer from a document they could not have opened themselves.
Agents that do the work. 16 connected services and 198 tools out of the box — plus any HTTP API, eight database engines and your own MCP servers.
Governed by default. A tool either runs on its own or stops for a person, and every run lands in an audit trail.

Explore the App Start free

For the team that wants working assistants this afternoon, without handing its documents to something it does not control.

The library underneath

Context engine

Permission-aware retrieval as a library over your own PostgreSQL. Full-text, trigram, vector and — where you enable it — graph legs, fused into one ranking, with all of them compiled through a single scope predicate. A document the caller may not open is never retrieved for them, by any route.

Two clients, one schema. pip install promptev-context-engine and npm install @promptev/context-engine — a corpus ingested by either is searchable by the other.
Your database, your models. PostgreSQL 14+ with vector, pg_trgm and unaccent; the engine creates them itself on first migration.
Apache-2.0. The part that enforces the permission boundary is the part you can read.

Explore the engine Read the guide

For the engineer building their own product who needs retrieval that respects who is asking — served from FastAPI, Flask or Django, from Hono, Express or Fastify, or over MCP.

How they fit together

One engine. Two ways in.

Every agent in the App reads through the Context engine — the same library you can install yourself. That is not a family resemblance: it is the same permission boundary, so what you audit in the open engine is what decides an answer in the App.

01 · The App

Build without writing code

Connect a source, describe an agent in plain words, decide what it may do on its own and what it must ask about, then put it in Slack, WhatsApp, a portal or your website. The engine runs underneath and you never touch it.

WorkspacesProjectsAgentsChannelsApprovals

02 · The engine, alone

Build with your own code

Install the Python or TypeScript client, point it at a PostgreSQL database you control, ingest your corpus, and search as a caller. No Promptev account is involved, and no data leaves your infrastructure.

Python 3.12+Node 22+PostgreSQL 14+Apache-2.0

03 · Both at once

Meet in the middle

Each project in the App is itself an MCP server at its own URL, so your own tools — an editor agent, a service you wrote — get that project's knowledge and tools with the caller's permissions intact.

MCP serverREST APIScoped keysWebhooks

Which one do I need

Start with the question you are answering.

Most teams start with the App and never install anything. Engineers building a product of their own start with the engine. Neither choice closes the other off.

The same question, answered for each product.
	Promptev App	Context engine
What it is	A product your whole team opens	A library your code imports
Who uses it	Anyone — owners, admins, editors, viewers	Developers, in Python or TypeScript
First step	Connect a source and describe an agent	`pip install` or `npm install`, then point it at a database
Where it runs	Managed by us, in your cloud, or on-premises	Wherever your PostgreSQL is
Retrieval	Built in — the engine, with the caller's permissions	The whole point — one scope predicate across every leg
Agents, tools, approvals	Included: 198 tools, orchestrations, approvals, audit	Not its job — you build that part
Channels	Slack, WhatsApp, portal, embeddable widget, webhook, MCP	Whatever you serve it from
Price	From $5 · credits, no seats	Apache-2.0 · free

The honest overlap. If you want permission-aware answers and nothing else, the engine is enough. If you want people who do not write code to build and run agents on that same boundary — with connected tools, approvals and an audit trail — that is the App.

Start here

Connect one source. Ask one question. See who gets what.

2,500 free credits, no card. Or install the engine and read the query it builds.

Start free Engine on PyPI Engine on npm

2,500 free credits · No card required · No subscription

Start Free →Book a Demo