for
Our approach to AI

Bring the AI you prefer. Give it a safe way to work in .

We do not ask an association to bet its future on one AI company or one model. keeps the iMIS® knowledge and working methods in one place, while the AI app around it can develop and change.

Request a demoStart the conversation

Choose the workhorse that fits your people and your rules.

Some teams prefer a friendly desktop app. Technical teams may prefer a command line. Some organisations are comfortable with a hosted model; others want a model on equipment they control. AgentZ is designed to keep the iMIS part consistent across those choices.

Choose the AI app and model that fit the work.

A model supplies the reasoning. An agent app supplies the working space, tools, and MCP connection. AgentZ supplies the safe working methods for iMIS. Select a logo to visit that product’s website. Use the setup links below when you are ready to connect AgentZ.

Examples only. These marks belong to their owners and do not imply endorsement. A model or chat product is not automatically an AgentZ client. The chosen agent app must support MCP, and the complete setup must pass the AgentZ connection and security checks.

Chat and agent work are not the same

The model thinks. The app carries the job. knows how to do the work in .

The same Claude, Gemini, Grok, or GPT model can sit inside a simple chat product or a full agent app. The model name alone does not tell you what it can do.

1 · Conversation

Chatbot-style AI

A chatbot is mainly a place to ask a question and receive an answer. It may read an uploaded file or search the web, but the normal result is still a reply in the conversation.

  • Useful for ideas, drafts, summaries, and questions.
  • Usually stops when it gives the answer.
  • Does not gain iMIS access because it uses the same model.
2 · Working environment

Agent app or harness

An agent app surrounds the model with the job itself. It can keep the instructions, files, plan, progress, tools, and earlier results together. When supported, it can use the terminal, browser, or computer, call MCP tools, inspect what happened, and continue. Codex, Claude Code and Desktop, OpenCode, Google Antigravity, Grok Build, Hermes Agent, OpenClaw, and Cursor are examples.

  • Works through several connected steps.
  • Can stop for approval and then resume.
  • Uses MCP to find approved outside tools.
3 · iMIS working suite

is the iMIS® working layer used by the agent app. It provides the iMIS® knowledge, maintained operations, signed-in access, evidence, approval points, recovery steps, and final checks needed to turn intent into permitted action.

  • Understands association work and how iMIS records connect.
  • Uses the person’s real iMIS permissions.
  • Checks the result instead of stopping at a plausible reply.
Three-panel comparison of generic Claude use, a direct do-it-yourself MCP connection to iMIS, and the governed AgentZ route
Three ways to connect AI and iMIS. The model is only one part of the answer. AgentZ adds the iMIS knowledge, maintained working methods, checks, and recovery steps that let the AI take on sustained operational work.
Four useful terms

What the technical words mean in normal work.

You do not need these words to use . They help when you compare products or discuss security.

AI model
The part that reads, reasons, and writes. Claude, Gemini, Grok, and GPT are model families. Different models are better at different work, and they can have different costs and data terms.
AI app
The app around the model. People sometimes call this a workbench or harness. It is more than a chatbot because it can keep the conversation, files, instructions, tools, plans, and progress together for a long job. Codex, Claude Code and Desktop, OpenCode, Google Antigravity, Grok Build, Hermes Agent, OpenClaw, and Cursor are examples.
The standard connection: MCP
MCP means Model Context Protocol. Think of it as a common connection point. It lets an AI app find and use AgentZ without every AI company needing a different custom link.
AI on your own equipment
The technical term is local inference. The AI model runs on equipment controlled by the organisation instead of using a public online model service. This can suit some privacy or cost needs, but it still needs access rules, testing, and human review.
Different ways to work

The interface can fit the person doing the job.

These choices change the working experience. They do not change who can see or change something in iMIS®.

Desktop app

A visual place to talk through the job, bring in files, see results, and approve important steps. This is often the easiest route for everyday staff.

Command line

A text-based place for technical users who want files, repeatable instructions, code, and longer automated work close together.

Online or local model

The model can be an online service or, when the chosen workbench supports it, run on equipment controlled by the organisation.

Connect the app your team chooses

Use the product’s own MCP setup guide.

These links explain where to add an MCP connection in each product. The person who manages the app can follow the guide and use the connection details supplied during setup.

OpenAI

Codex

Codex can use an MCP connection from the desktop app, command line, or editor extension. These local Codex clients share the same server settings.

Set up MCP in Codex
Open source

OpenCode

OpenCode can connect to a local or remote MCP server. Its guide shows the settings, sign-in, and connection checks.

Set up MCP in OpenCode
Google

Antigravity

Antigravity has an MCP Store in the app and an MCP manager in its command-line view. It can also accept a custom server configuration.

Set up MCP in Antigravity
xAI

Grok Build

Grok Build keeps MCP server settings with its other user or project settings. Its guide explains the supported connection types and checks.

Set up MCP in Grok Build
Cursor

Cursor Agent

Cursor can add a server from its MCP collection or from an MCP settings file. Its guide also explains authentication and safe use.

Set up MCP in Cursor
What does not change

A different AI does not get different rights in .

The AI can help carry out the work. It does not become the person who owns the decision.

A known personAgentZ works as the person signed in to iMIS, with that person’s access.
Show the recordsAgentZ explains which iMIS information supports the answer.
Ask at the right timeImportant changes wait for the person to make the decision.
Check the resultAgentZ reads the result back and records what happened.
Choice still needs testing. Not every AI app or local model works with AgentZ automatically. It must support the standard connection, meet the organisation’s security needs, and be tested for the work the team wants to do.
Continue

Read the other parts of the AI guide.

Each page answers a different question without requiring technical knowledge.

Our approach to AI

How to choose an AI app or model without rebuilding the connection to iMIS.

Read the approach

How works

How turns a request into checked work in iMIS® while the person stays in charge.

See how it works
Next step

Show us the work your team actually needs to get done.

Bring something real: member service, renewals, an upcoming event, reporting, data quality, or a staff dashboard. The demo is where we work out the fit, and the rest follows from there.

Request demoClient download
iFINITY

is the AI teammate for associations running iMIS® — built by iFINITY, a Premier Authorised iMIS® Solution Provider.

Trusted across the communityRecognised relationships for iMIS® delivery and the membership sector.
Authorised iMIS Solution ProviderMemberWise Recognised Supplier 2026Membership World Vetted Partner 2023