Skip to content

Agent Mode

Your digital colleague

An IntraGPT agent is a digital colleague that you train yourself, equip with tools and connectors, and then let work independently. Not a stand-alone chatbot, but a colleague with a job title, expertise, its own memory and access to exactly the systems you grant it.

Book a free AI consultation View the platform
  • Live Chromium browser the agent drives itself
  • Word, PDF, Excel and QR codes from one prompt
  • Custom agents per team with their own tools and MCP integrations
IntraGPT Agent Mode with a live browser window inside the chat

They already chose IntraGPT

What an agent can do

One instruction, multiple steps. The agent keeps working on its own until the task is done.

Knowledge and lookup

  • Knowledge base (RAG): searches the company knowledge base and answers with source references.
  • Knowledge articles: consults internal articles and prompt templates you wrote yourself.
  • Web search (Google): searches the web live with sources, without opening a browser.
  • Deep search: agentic, in-depth research across multiple sources.
  • Chat with documents: talk directly to uploaded documents.

Documents and files

  • Word (.docx): documents with text, tables, formatting and your house style.
  • Filling in Word templates: completes your own templates (quote, contract, letter) automatically.
  • Excel (.xlsx): styled spreadsheets with tables and data.
  • PDF: print-ready professional PDF documents.
  • QR codes: high-resolution QR codes for a URL, text, contact details or wifi.

Calculation and data

  • Calculator and statistics: exact calculations such as sum, average and standard deviation.
  • Data analysis: summarise, group, filter and sort CSV files and tables without code.
  • Unit conversion: length, weight, temperature, volume, time, storage, speed.
  • Querying JSON: pulls specific values out of JSON data.

Email and communication

  • Sending email: sends emails with attachments on your behalf.
  • Reading and handling a mailbox: reads, searches and answers mail from a connected mailbox.
  • Any provider: connect a mailbox over IMAP/SMTP. New mail is fetched and processed automatically.
  • Every incoming message starts the agent, or it picks up mail itself when it sees the need.
  • Draft mode: you approve the reply. Autonomous mode: it replies itself in the same thread.
  • Passwords are stored encrypted. Only mail from the moment of connecting, not your entire archive.

Operating systems and tools

  • 90+ ready-made connectors: retrieves and updates data in the tools your company already uses.
  • HTTP requests and webhooks: calls external APIs and triggers external systems.
  • MCP servers: Model Context Protocol servers become available to the agent automatically.
  • Managed browser: drives a real Chromium browser for public websites.
  • Chrome extension: drives your own Chrome with your logged-in sessions.

Memory

Long-term memory: remembers stable facts and preferences, kept separate per user and per agent.

The training

An agent you train and certify

An IntraGPT agent is not just configured, it is trained. From skills to a verification score before it goes live.

  1. 01

    Skills

    Teach the agent concrete competences with explanations and examples.

  2. 02

    Knowledge

    Add curated knowledge that the agent has at hand.

  3. 03

    Learning material

    Feed in documents and URLs; the system distils skills and knowledge from them.

  4. 04

    Exams

    Scenario-based tests check whether the agent really masters the material.

  5. 05

    Verification

    Verification status and score before going live. You know it is ready.

Autonomy: work that happens by itself

From guided suggestions to fully autonomous workflows that schedule themselves.

Autonomy level

Choose per agent: guided (you approve) or autonomous (it handles things itself).

Self-scheduling

Let the agent start itself on a schedule, with an adjustable daily maximum.

Smart Tasks

Simple scheduled tasks: one agent, one fixed assignment.

AI Workflows

Visual multi-step pipelines with triggers, LLM steps, tools and branches.

Connectors

90+ integrations

The agent retrieves data and updates it in the systems you already use.

CRM HubSpot, Salesforce, Pipedrive, Teamleader, Zoho CRM, Close, Freshsales, Copernica
Accounting and finance (NL) Exact Online, Moneybird, e-Boekhouden, Twinfield, SnelStart, AFAS
Payments Stripe, Mollie, PayPal, Pay.nl, MultiSafepay, Chargebee, Xero, QuickBooks
E-commerce and fulfilment Shopify, Bol.com, Picqer, Channable, Sendcloud, WordPress
Marketing and email Mailchimp, Klaviyo, Brevo, ActiveCampaign, SendGrid, Postmark, Resend, ConvertKit, Spotler
Projects and tasks Asana, Jira, Trello, ClickUp, Monday, Linear, Wrike, Todoist
Communication and support Slack, Teams, Discord, Twilio, MessageBird, WhatsApp Business, Trengo, Zoom, Intercom, Zendesk, Freshdesk, TopDesk
HR and payroll BambooHR, Personio, Nmbrs, Loket.nl, Simplicate
Google and Microsoft Gmail, Google Agenda/Drive/Sheets/Analytics, Outlook, OneDrive, Microsoft Agenda/Teams
Content and data Notion, Airtable, Contentful, Dropbox, YouTube, Algolia, Ahrefs
Analytics Google Analytics, Mixpanel, Amplitude, Segment
Developer GitHub, GitLab, Sentry, Datadog, Vercel
Social Instagram, LinkedIn, X (Twitter)
Automation Zapier, Make, or a full API of your own

Everything around it

Everything that makes an agent usable in daily practice: templates, workflows, knowledge, document management and the channels it is reachable on.

Word templates

Upload your own templates with variables. The agent fills quotes, contracts and letters automatically with the right data from your systems.

AI Workflows and Smart Tasks

Build multi-step pipelines visually, with triggers, LLM steps, tool calls and branches. Every run is logged step by step. Smart Tasks are the simple counterpart: one agent, one scheduled assignment.

Knowledge in four flavours

Knowledge base (RAG): vector search across your company knowledge with source references. Knowledge articles: hand-written internal knowledge and prompt templates. External knowledge bases: outside sources synchronised on a schedule. API knowledge sources: knowledge fetched periodically from external APIs.

DMS document management

Documents with categories, types and departments. Approval workflows, mandatory read receipts, feedback and a complete audit trail.

Social Manager

Manage social posts and the image assets that go with them from the same platform.

API Portal

Programmatic access to the platform for your own integrations and developers.

Reachable everywhere

The same agent through whichever channel fits: the web dashboard as the central chat and administration environment, the mobile app for iOS, iPadOS, macOS and Android, WhatsApp, Telegram, email as a fully fledged working channel, the Chrome extension that acts inside your own browser, and the API for your own applications.

Multilingual, secure and walled off per organisation

An AI that doesn’t just answer, it acts. On your own language model, inside your own environment, without sharing data with external parties.

Multi-tenant isolation

Every organisation strictly separated. An agent never sees data from another organisation.

Encrypted secrets

Passwords and API keys stored encrypted, never in plain text.

White label

Your own theme and house style per organisation.

Multilingual portal

Available in Dutch, English and German.

Walled off per agent

You decide which tools, connectors and knowledge each agent may use.

What an agent deliberately does not do

Fair is fair. A good digital colleague knows its limits.

  • Only what you allow. No connector means no access to that system.
  • Knowledge is only as good as what you supply. Without a source it invents nothing, it reports that the source is missing.
  • It acts within the limits you set. In guided mode: no irreversible actions without approval.
  • Mailboxes: only mail from the moment of connecting. Never your entire historical inbox.
  • No black box. Answers come with their source and workflow runs can be traced step by step.
  • It stays inside your organisation. Strict data isolation between tenants.
  • A language model remains a language model. Hence the training, the exams, verification and optional human approval.

Ready for your first digital colleague?

We show you live how an agent is trained, connected and put to work inside your organisation.

Book a free AI consultation Contact us
A reply within one working day, no obligations. Enschede and Amsterdam.