NEWPromptev MCP Server: one URL for every tool, every agent and your team’s knowledge. Connect Claude Code, Cursor, or any MCP client.→

The sovereign agentic AI platform

Sovereign AI agents for your business.

promptev is the AI agent platform for business. Build AI agents without code on your company's knowledge and tools. They answer in every channel and write the reports your team shares, while your data never leaves your control.

Start free with 2,500 credits
Up to 83% lower AI costSOC 2 Type IIYour own model keysYour database, optionalNo seats, no plans
promptev consolelive recording · demo store

Lower AI cost

Cut your AI bill by up to 83%.

Most AI agents get more expensive as your data grows. promptev agents answer from large data at a fraction of the tokens, on your own model keys.

Start freeSee the numbers →
promptev consolesample data
The console's saved panel: the amount saved, the amount paid, and the share savedThe console's saved panel: the amount saved, the amount paid, and the share saved
Your console shows what you paid, and what you saved.
83%

lower cost per answer

78%

fewer tokens per answer

100,000

records in one answer

$0.009

cost per answer on 100,000 records

1,000,000

records in our largest test

38

model providers, on your own keys

No-code AI agent builder

Build AI agents in 60 seconds. Watch them work.

0:00

1 · Connect

Plug in everything your company knows.

Google Drive, SharePoint, Slack, Jira, your databases, any HTTP API or MCP server. Sign in once and each source's own sharing comes with it.

  • Pick folders, files and tables
  • Kept in sync automatically
0:12

2 · Build

Describe the job. No code.

Give the agent instructions, knowledge and tools. Choose who may use it and which actions need a person to approve first.

  • Runs on your own model keys: OpenAI, Anthropic, Google, Mistral
  • Chain agents into flows with AI agent orchestration
  • Deploy in one click
0:24

3 · Work

Ask a question. Watch every step.

The agent plans, searches your documents, queries your data and writes the answer, with a source for every number. When it wants to act, like emailing the board, it stops and asks first.

  • Every step visible while it runs
  • Citations you can open in one click
  • Approvals in the chat, in Slack or by email
0:36

4 · Deliver

Not just answers. Deliverables.

Reports, spreadsheets and decks, saved as artifacts with every version kept. Share them with named colleagues or anyone with the link.

  • An AI report generator and AI document generator that knows your data
  • Hand an artifact back to any agent to update it
0:48

5 · Govern

Each person sees only what they may.

AI governance built in: two people ask the same question and get two correct answers. Every question, answer and approval is on the record.

  • Permissions checked inside the search itself
  • Emails, phone and card numbers redacted before any model sees them
  • Every question, answer and approval on the audit record
Connect· promptev
Google DriveSharePointSlackGmailOutlookJiraConfluenceDropboxOneDriveGoogle SheetsWhatsAppMCP servers
promptevCONTEXT ENGINE
12 sources connected · permissions kept
FA
Finance analystNew agent · Finance project

Answer questions on revenue, costs and forecasts. Cite every number. Write board-ready reports.

Finance folderQ3 forecastsales_db
Send emailPost to SlackCreate issue

Finance team

Before sending email

Deploy Live in Slack, WhatsApp, email and the web
● Live
FAFinance analyst · asked by Aisha
Summarise Q3 revenue against plan, write it up as a board report and email it to the board.
  1. Searching Google Drive Finance folder
  2. Reading Q3 forecast.xlsx
  3. Querying sales_db.revenue
  4. Writing the board report

Q3 revenue closed at $4.82M, 6% above plan, led by Enterprise (+18%). The board report is ready.

Q3 forecast.xlsxsales_db.revenueQ3 board report · v1
Send "Q3 board report" to [email protected]?gmail.send_email · needs approval
ApproveDeny✓ Approved by Aisha · sent · logged
DOCQ3 board report

Q3 revenue: 6% above plan

JulAugSepPlan
Version 1
Share "Q3 board report"
SYou Owner
AAisha Khan Can view
MMarco Diaz Can view
🔒 Only people invited
Shared · every version kept
Both askWhat is our payroll cost this quarter?
NNadiaFinance director
FinanceHR payroll

$1.94M, up 3% on Q2. Engineering is 61% of it.

Answered from 3 sources
TTomContractor, Support
Support docsHR payroll

That's in documents you don't have access to. I can help with support policies.

Stayed inside Tom's access
Redacted in place, before the model

Refund Sara Ali at [email protected][email], card 4242 4242 4242 4242[card ••4242], phone +971 50 123 4567[phone].

The model only ever sees the masks
Audit record
14:02Nadia asked Finance analyst · 3 sources read
14:02Tom asked Finance analyst · HR payroll withheld
14:05Aisha approved gmail.send_email · board report sent
14:073 personal details redacted before the model

That's promptev. Now try it on your own data.

Connect a source and talk to your first agent in minutes.

$0 to start2,500 free creditsNo cardEvery feature included
Start freeBook a demo

Connectors

Connect your AI agents to everything your company knows, in one sign-in.

Connect the apps your team already uses. Knowledge stays in sync and follows each source's own sharing, so nobody sees more through an agent than they could see in the app.

Google DriveFiles and foldersGmailMail and draftsGoogle CalendarEvents and free timeGoogle DocsDocumentsGoogle SheetsSpreadsheetsGoogle SlidesPresentationsJiraIssues and projectsConfluenceSpaces and pagesDatabasesPostgres, MySQL, SQL Server
WhatsAppBusiness messagingDropboxFiles and foldersModel Context ProtocolAny MCP serverMicrosoft OutlookMail and calendarMicrosoft OneDriveFiles and foldersMicrosoft SharePointSites and librariesAzure DevOpsWork items and reposSlackChannels and threadsHTTP APIsAny REST endpointUploadsPDF, Word, Excel and more
Any database

Postgres, MySQL, SQL Server and more, limited to the tables you pick.

Any HTTP API

Describe an endpoint once and every agent can call it, with approvals where it matters.

Any MCP server

Bring tools from the MCP ecosystem, and serve your own agents as an MCP server too.

Everywhere your team works

AI agents for business, in every channel.

Build it once. Your people reach it where they already are, and it knows who each of them is.

Slack

@finance what did we spend on cloud in August?
$38,420, 9% under budget. Breakdown attached.
Channels, threads and DMs

WhatsApp

Is order #2231 shipped?
Yes, it left Dubai this morning. Arrives Thursday.
A WhatsApp AI agent on your business number

Your website

Do you ship to Canada?
We do, in 5 to 7 days. Want the rates?
One script tag, any site

iOS, Android, web

Draft the onboarding plan for Sara
Done: Onboarding plan, version 1.
The promptev app

MCP server, SDKs and API

An MCP server for your agents, in Claude, ChatGPT and Cursor.

promptev is an MCP server you sign in to. Add one URL to Claude, ChatGPT, Cursor or any MCP client and sign in with your promptev account: there is no key to copy. The app then works as you. It builds and publishes agents, searches your company knowledge, runs your tools and calls your agents, with your own permissions, approvals and audit.

Your AI app · signed in to promptev as you
Ask our finance agent what Q3 revenue was, then open a Jira issue for the Self-serve gap.
Asked the Finance analyst→ $4.82M, 6% above plan · 3 sources
Opened a Jira issue→ waiting for approval · approved by Aisha · FIN-212

Q3 revenue was $4.82M, 6% above plan. I opened FIN-212 for the Self-serve shortfall after Aisha approved it.

  • Build and publish agents from your AI app
  • Ask any agent and get an answer with its sources
  • Search your company knowledge
  • Run your connected tools, with approvals
  • Connect apps and add files, through a link only you can open
{ "mcpServers": { "promptev": { "url": "https://api.promptev.ai/mcp" } } }

One address for your whole workspace. Sign in when your app asks. How it works →

npm i @promptev/clientpip install promptevRead the docs →

ClaudeChatGPTCursorVS CodeAny MCP clientNode.jsPython

Context engineering, done for you

The context engine: answers that are right, and private.

A context engine decides what an AI reads before it answers. It is context engineering done for you: the right facts from your company, for the person asking, and nothing else.

Google Drivetheir sharing
SharePointtheir sharing
Sales databasepicked tables
Confluenceyour groups
promptev context engine

Find. Scope. Protect. Prove.

  1. Find by meaning and by exact words, across files, rows and the links between them.
  2. Scope to the person inside the search itself, so the model never reads what they can't open.
  3. Protect sensitive fields by redacting them before any model sees them.
  4. Prove each claim with the file, row or page it came from.
Answer for Nadia$1.94M, up 3% on Q2.Payroll Q3.xlsx · row 42
Board pack Sept.pdf · p. 7Cited · scoped · logged
promptev console · a real answer
A promptev agent answers a question in four points and names the file it usedA promptev agent answers a question in four points and names the file it used
An agent answers from your own knowledge, and names its source.
0

of your documents used to train models.

1

check, inside the search, decides what each person's AI may read.

100%

of actions on the record: who asked, what ran, what it cost.

Open

source engine, Apache 2.0. Read the code that guards your data.

Sovereign AI

Sovereign AI: your keys, your database, your cloud.

Sovereign AI means you decide where every part runs. Your agents run on your own model keys from day one. Start on our managed cloud in minutes, and bring your database, storage and deployment home when you are ready.

Models

Your own keys for OpenAI, Anthropic, Google, Mistral, DeepSeek or your own endpoint. Billed by your provider, never marked up by us.

Your keys
Knowledge

Drive, SharePoint, Dropbox, Confluence, databases and uploads, following each source's own permissions.

Stays yours
Database

Your agents, conversations and knowledge index in our managed database, or in your own.

Bring your own
Storage

Files and artifacts in our storage, or in your own bucket.

Your bucket, optional
Deployment

Managed cloud, or self-hosted AI agents in your own cloud account or fully on-premise.

You choose

Compare

An n8n alternative built for AI agents

Looking for an n8n alternative or a Zapier alternative? Automation tools move data between apps. promptev gives each person an agent that knows your company, and only what they may see.

promptev vs Gleanpromptev vs Microsoft Copilotpromptev vs ChatGPT Enterprisepromptev vs LangChain and LangGraphpromptev vs n8n and Zapierpromptev vs building your own

Pricing

Start at $0. Pay only for what agents do.

No seats and no plans. Every feature is in every workspace. You buy credits when you need them, and model calls run on your own keys.

Start free$02,500 credits on sign-up, worth $12.50
  • No credit card
  • Every feature, nothing held back
  • SSO, approvals and audit included
Start free
Pay as you go$0.005 per credit$5 minimum top-up: 1,000 credits
  • Credits never expire
  • Optional auto top-up
  • Add people for free, pay only for use
Get started
EnterpriseCustomEnterprise AI agents in your cloud or on-premise
  • An annual licence, not metered credits
  • Onboarding and support
  • Data kept in one jurisdiction
Talk to us
1 credit = 1 AI action. Most questions use 2 to 5 credits. Embedding is always free.See full pricing →

Questions

Answers first

How much can promptev lower my AI cost?

In our tests on large data, promptev agents answered at up to 83% lower cost and with up to 78% fewer tokens than before our October 2026 release. The saving is largest on questions that need a lot of data. promptev runs on your own model keys, so fewer tokens is a lower bill from your model provider.

What is a sovereign AI agent?

An AI agent whose models, data and deployment stay under your control. promptev runs in our managed cloud, in your own cloud account or on-premise, runs on your own model keys, can use your own database, and only ever reads what the person asking is allowed to see.

How much does promptev cost?

You start at $0 with 2,500 free credits and no card. After that you pay as you go at $0.005 a credit, with a $5 minimum top-up, and credits never expire. There are no seats or plans. Model calls run on your own provider keys. Enterprise deployments in your own cloud are an annual licence.

Can I use promptev as an MCP server, or from code?

Yes, two ways. Add https://api.promptev.ai/mcp to Claude, Claude Code, ChatGPT, Cursor or any MCP client, sign in with your promptev account, and build and publish agents as yourself, with your own permissions. And every project is an MCP server too: add its URL and a project key to call your agents, search your knowledge and run tools, with the same permissions and approvals. Developers can also use the JavaScript and Python SDKs or the REST API.

What is a context engine?

A context engine chooses what an AI model reads before it answers: it finds the relevant facts across your documents and systems, keeps only what the person asking is allowed to see, and cites where each fact came from. promptev's context engine is open source.

Is promptev private AI for business?

Yes. Each answer only uses what the person asking may see, your data is never used for training, sensitive fields can be redacted, and you can run everything in your own cloud.

Can I self-host promptev AI agents?

Yes. Enterprise customers run self-hosted AI agents in their own cloud account or fully on-premise, with their own database, storage and model keys.

Is promptev a no-code AI agent builder?

Yes. You build an agent by describing its job, attaching knowledge and tools, and choosing who can use it. Developers can also reach every agent through the API and as an MCP server.

How is promptev different from n8n or Zapier?

n8n and Zapier automate workflows between apps. promptev gives your people AI agents that answer from company knowledge, take actions with approvals, and respect each person's permissions.

Can the agents write documents and reports?

Yes. Agents create documents, spreadsheets and decks as artifacts. Each keeps its versions and can be shared with colleagues or by public link.

Build your first AI agent today, for free.

Start at $0 with 2,500 free credits. Then pay as you go from $5.

Start free with 2,500 creditsBook a demo
No cardNo seatsYour own model keysSOC 2 Type II
2,500 free credits · No card required · No subscription
Start Free →Book a Demo