Skip to content
Menu

AgentGate

Your systems, your data, now using AI.No migrating. No exposing. Secure.

AgentGate turns your systems into secure business tools for ChatGPT, Claude and any AI agent, and governs the models they use.

  • No inbound ports
  • Permissions per tool
  • Every query audited

Your team already uses AI, just not with your company's data

Risk

Data copied by hand

Employees paste information from your systems into a chat, with no control or record.

Gap

Systems AI cannot reach

Databases and in-house systems are left out of standard connectors.

Cost

Slow, expensive integrations

A custom project takes months and leaves code that someone has to maintain.

Your network stays closed: the connector reaches out to AgentGate

The connector is installed in your network and it is the one that starts the connection. Your systems never receive connections from the Internet.

Your team

ChatGPT, Claude or other assistants

AgentGate

Permissions, rules and audit

Connector

In your network, no inbound ports

Your systems

Databases and APIs

The connector starts the connection: outbound, encrypted and certificate-based (mTLS).

  1. 01

    Install the connector

    A Docker container in your network, with a guided installer.

  2. 02

    Define tools

    Business questions about your systems, with the exact fields that are returned.

  3. 03

    Assign access

    Each tool only for the people and groups it belongs to.

Three things in one platform

Connect

Your databases and APIs become tools that any compatible assistant can use.

See how it connects →

Govern

Who can use each tool, what it can change and what was recorded.

See the controls →

Control the models

A single access point to AI models, with budgets, per-person keys and security steps.

See the model gateway →

Secure by design, not by configuration

01

No inbound ports

The connector only opens an outbound, encrypted connection.

02

Credentials stay in your network

With the connector, your systems' passwords never leave your network.

03

Never free access

Only predefined, approved and signed tools are executed.

04

Denied by default

Each tool is assigned to specific roles or groups of your organization.

05

Changes are confirmed

Every tool that writes asks for confirmation before it runs.

06

Full audit trail

Who queried what, from which client and with what result.

Direct mode (optional): if you prefer not to install anything, the platform can connect on its own to reachable systems; in that case the credential is stored encrypted in AgentGate.

It also governs the models your agents use

All traffic to the models goes through a single point inside your network. You know who spends how much, you set limits and you keep sensitive data from reaching an external provider.

  • Compatible with OpenAI and Anthropic
  • Budgets per organization, group and person
  • Virtual keys with one holder each
  • Personal data masked before the model
  • Which model each person uses, in one place
  • Which MCP servers each person uses

Plans

Start on your own or with your team, or book a demo for your company.

Individuals and small teams

You sign up and subscribe in a few minutes.

Personal

For individual use

$19 / month

  • 1 connector
  • Up to 20 tools
  • Up to 2 modules
  • AI assistant connection
  • Read-only by default
  • Change confirmation
  • Model gateway
Get this plan

Team

For small teams

$79 / month

  • Up to 5 active users
  • 1 connector
  • Up to 20 tools
  • Up to 2 modules
  • AI assistant connection
  • Connector in your network
  • Every query audited
  • Model gateway
Get this plan

Companies

We start with a demo and a pilot on your real case.

Starter

To start with one system

$240 / month

  • Up to 25 active users
  • 1 connector
  • Up to 50 tools
  • Up to 5 modules
  • 90-day audit trail
  • Connector in your network
  • Every query audited
  • Change confirmation
Get this plan

Business

Recommended

For several systems and teams

$690 / month

  • Up to 100 active users
  • 3 connectors
  • Unlimited tools
  • Unlimited modules
  • 365-day audit trail
  • Connector in your network
  • Every query audited
  • Change confirmation
  • Corporate sign-in
  • Model gateway
Get this plan

Enterprise

For the whole organization

Contact us

  • Unlimited connectors
  • Custom audit retention
  • Connector in your network
  • Every query audited
  • Corporate sign-in
  • Model gateway
  • SLA and priority support
  • Dedicated region
  • Custom data retention
Get this plan

Prices are expressed in US dollars (US$). Reference prices, subject to confirmation.

Add-ons

Ask us about these add-ons.

Automated agents

Agents that work on their own with their own identity and quotas.

Industrial sites

Connection of plants and industrial systems.

Frequently asked questions

What is AgentGate?

A platform that turns your databases and APIs into secure tools for AI assistants such as ChatGPT or Claude, and also controls access and model spend.

Do I have to open ports in my network?

No. The connector is installed in your network and it starts an outbound, encrypted connection to AgentGate. There is also an optional direct mode that installs nothing.

Do my systems' credentials leave my network?

With the connector, no: they stay in your network. In the optional direct mode, the credential is stored encrypted in AgentGate.

Can the AI change my data?

Only if you allow it. Every connector is read-only by default, and every tool that writes asks for confirmation before it runs.

Which assistants does it work with?

Any MCP-compatible client. There are connection guides for Claude, ChatGPT and OpenCode.

What does the model gateway do?

It is a single access point to AI models: compatible with OpenAI and Anthropic, with budgets, per-person keys and steps such as personal data masking.

Which systems do you want to connect to AI?

Tell us your case and we will show you how it would look with AgentGate.

Talk to us