Yes. Halo MCP uses OAuth 2.0 for authentication. You can create separate Tool Collections with different permission levels — for example, a user-facing collection that can only raise tickets, and an agent-facing collection with broader capabilities. Each collection has its own unique endpoint URL and access controls.

Do we need to build anything custom to use Halo MCP?

No custom development is required. Halo MCP is a built-in feature enabled at Configuration > AI. Allied ESM can scope and configure it for you — enabling the endpoint, building the Tool Collections, setting up authentication, connecting your preferred AI assistant and testing the integration. That configuration work is scoped and quoted as part of your project.

What is the difference between Halo MCP and the Halo Virtual Agent?

The Halo Virtual Agent is a chat interface built inside the Halo self-service portal. MCP is an open protocol endpoint that lets external AI assistants — tools your team already uses — connect to Halo. They are complementary: the Virtual Agent handles in-portal conversations, while MCP lets you bring Halo capabilities into whatever AI tool your organisation prefers.

Can the AI see all of our Halo data through MCP?

No. Access is scoped three ways: the AI can only call the functions on the tool list you configure, it can only read the data behind the specific reports you expose to it from Halo's reporting suite, and each Tool Collection has its own endpoint, OAuth application and audience so different groups never share permissions. The in-Halo chat terminal currently inherits permissions from the account the integration authenticates as rather than the individual agent, so Halo recommends restricting that deployment to admin agents.

Halo's Latest AI — MCP

Halo MCP —
Your AI Assistant, Inside Your Service Desk.

Halo's Model Context Protocol (MCP) endpoint lets any AI assistant — ChatGPT, Claude, Copilot — discover and call your service desk functions directly. Raise tickets, search knowledge, log time, action requests. All through conversation.

Talk to an MCP Specialist

Open standard / Works with any AI tool

No code / Built into Halo

OAuth 2.0 / Secure by default

User via AI Assistant

ChatGPT · Claude · Copilot

"Log a P2 ticket — my VPN keeps dropping on the London network."

Done. Ticket #4821 raised, Priority 2, assigned to Network Team. Three similar past incidents found — want me to share the resolution steps?

"Yes, and log 30 mins against it."

30 minutes logged. Here are the top 3 resolution steps from ticket #3107...

Ticket raised KB searched Time logged

MCP explained

What is MCP, and Why Does It Matter?

Most AI tools are disconnected from your systems — they only know what you paste in. MCP (Model Context Protocol) is an open standard that gives AI a secure, live channel into your tools. Instead of describing a ticket, the AI reads it. Instead of telling it to update a record, it does it.

Your tools

Halo, email, docs

→

MCP layer

Secure protocol

→

AI assistant

Claude · GPT · Copilot

→

Action taken

In your live system

Halo has built an MCP endpoint directly into the platform. Enable it, connect your AI tool of choice, and your service desk becomes something any AI assistant can talk to. No middleware. No custom development. No third-party connectors.

Without MCP

"What priority should this ticket be?"

Agent pastes ticket text into an AI chat window

AI responds: "I'd suggest Priority 2 — here's why…"

Agent switches back to Halo and navigates to the ticket

Agent manually sets the priority field and saves

4 steps · 2 applications · manual data entry

With Halo MCP

"What priority should this ticket be?"

AI reads the ticket directly from Halo, evaluates urgency and context, sets Priority 2, and confirms — all without leaving the conversation.

Ticket #4821 updated · Priority set to P2 · Done.

1 step · 1 conversation · zero manual input

What Your AI Can Do in Halo

Halo MCP exposes a set of ready-to-use tools that any connected AI assistant can discover and call. These are available out of the box — no configuration required to get started.

Get or Search Tickets

Pull back full ticket data or search across ticket content. The AI can summarise, compare, and suggest replies or resolutions based on what it finds.

Add a Note to a Ticket

The AI can write a private note to any ticket on behalf of the agent — useful for logging investigation steps, handoff notes, or AI-generated summaries.

Action a Ticket

Trigger any configured quick action in Halo — update status, send a template email, reassign, escalate — without the agent touching the interface.

Search Knowledge Base

AI-powered semantic search of your knowledge base. Returns resolution suggestions and relevant articles — and can send them directly to users.

Log Time

Quick time entry against any ticket. The AI can log time as part of a multi-step task — for example, raise ticket, investigate, then log 45 minutes — all in one conversation.

Get User Info

Retrieve Halo user records — contact details, organisation, open tickets. Gives the AI context about who it's helping before it takes any action.

Search Service Catalogue

AI search of your service catalogue — returns service names and links so users can be directed to the right request form without navigating the portal themselves.

Apply AI Suggestions

Trigger Halo's AI suggestions engine and apply the results — automatically setting category, priority, team, and other fields based on matched historical tickets.

Get AI Matched Articles & Tickets

Re-run AI matching for any ticket on demand, returning the most similar past tickets and relevant KB articles based on the current ticket content.

Get Assigned Tickets

Returns all tickets currently assigned to the logged-in agent. The AI can summarise workload, highlight urgency, and suggest what to tackle first.

Assign to Me

The AI can reassign an open ticket to the currently authenticated agent in a single step — no navigating to the ticket required.

Get One Article

Retrieve and summarise a specific KB article by ID. The AI can extract the key steps, adapt the language for the user, and send it directly in the conversation.

Plus custom tools: Log a specific ticket type · Execute a Runbook · Get report data · Trigger an AI Ability

Tool Collections: One Endpoint Per Use Case

Create multiple MCP endpoints, each with a different set of tools and a different permission level. A self-service endpoint for end users. A full-capability endpoint for agents. Each with its own URL and access controls.

User Functions

Self-service · Portal · External

Log a ticket (any type or template)
Search Knowledge Base
Get my tickets
Update a ticket
Search Service Catalogue
Get User Info
Get One Article
Execute a Runbook
Trigger an AI Ability

Agent Functions

Full access · Internal use · Agents only

Everything in User Functions
Get & Search all tickets
Action a ticket (all quick actions)
Add private note to a ticket
Apply AI Suggestions
Get matched articles & tickets
Assign to me
Get assigned tickets
Log time

Each Tool Collection generates its own secure MCP endpoint URL and OAuth application. Available from Halo v2.232+.

Inside or outside

The Chat Doesn't Have to Live Somewhere Else

Most MCP conversations happen in the AI tool your team already uses. But Halo can also host the conversation itself — a chat terminal on the Halo home screen for service desk managers and power users, working against the same endpoint.

An AI terminal on your Halo home screen

Turn it on and a chat panel sits in the corner of the Halo home screen — you decide how large. It already knows who is logged in and which team they run, so questions don't need explaining.

Asks answered in place. “How is my team performing this week?” is answered from live data, without opening a report.

Charts rendered into the conversation. Ask for the breakdown as a graph and the image comes back in the thread.

Reading and doing in one thread. The same conversation can pull the numbers and then act on them — raise, assign, note, log.

Outside Halo

The assistant your organisation has already standardised on — ChatGPT, Claude, Copilot. Best for end users and agents who live in chat.

Inside Halo

A terminal on the Halo home screen. Best for managers and power users who want answers about their own desk without leaving it.

Halo AI · Service Desk

How is my team performing this week?

148 tickets resolved this week, up 12% on last week. Average first response 22 minutes. Two tickets are outside SLA.

Tickets resolved — by agent

J. Whitfield42

A. Rahman36

S. Okonkwo31

M. Traynor24

D. Kelleher15

Ask about your data…

Illustration of the in-Halo AI terminal — not a product screenshot.

Governance

You Choose the Tools. You Also Choose the Data.

Handing an AI assistant your whole API is not a security model. Halo's approach is narrower on purpose: the AI gets a named list of functions, a named set of data, and an endpoint scoped to a named audience.

The tools it may call

Every function is an explicit entry on a list — raise a ticket, add a note, search the knowledge base, log time. Nothing outside that list is reachable, and you can add your own.

The data it may read

Rather than opening the database, you expose specific reports from Halo's reporting suite. The AI answers from those. If a report isn't on the list, the data behind it isn't in scope.

The people who may use it

Each Tool Collection gets its own endpoint URL and OAuth application — a narrow one for end users, a broader one for agents. Different audiences never share a set of permissions.

Worth knowing: the in-Halo chat deployment is newer than the endpoint itself, and permissions don't yet inherit from the individual agent's Halo profile — they come from the account the integration authenticates as. Halo's own guidance is to keep the in-app terminal to admin agents until per-agent inheritance ships. We'd say the same: start it with the people who already see everything.

What turning it on actually involves

Four steps, all inside Halo's configuration — no middleware and no custom development.

Turn MCP on

Configuration → AI. The MCP area lists the tools the AI may call, and lets you add your own.

Add the runbook

Halo ships an MCP integration runbook — built in on new instances, importable from Halo's online repository on existing ones.

Create the credentials

Generate an API application for a client ID and secret. That produces the bearer token the endpoint authenticates with.

Connect the assistant

Point your AI tool at the endpoint URL. Halo publishes setup guides for Microsoft Copilot and OpenAI, plus one for deploying it in Halo's own chatbot.

Allied ESM can scope and configure this on your instance — endpoints, tool collections, data scoping and the AI connection. Talk to us about what's involved →

Three Ways Organisations Use Halo MCP

From replacing portal navigation to giving agents a genuine AI co-pilot, MCP opens up use cases that weren't possible before.

The Conversational Service Desk

End users raise tickets, check status, and find answers through a chat interface they already use — without ever logging into the portal.

User messages their AI assistant: "My laptop screen has gone black after an update."

AI searches the KB — finds a known resolution from a past incident.

AI shares the fix. User confirms it didn't work — AI raises an Incident ticket automatically.

User receives ticket number and expected SLA — all without opening Halo.

The Agent AI Co-Pilot

Agents use an AI assistant alongside Halo — the AI handles the repetitive work so the agent can focus on resolution.

Agent asks AI: "Summarise my ticket queue and flag anything over SLA."

AI retrieves all assigned tickets, identifies two breaching SLA, summarises each.

Agent says "Assign the VPN ticket to me and add a note that I'm picking it up."

AI assigns the ticket and writes the note. Agent moves straight to resolution.

External AI Connects to Halo

From v2.236+, Halo can also connect outbound to third-party MCP servers — so Halo's own AI agents can use tools in external systems.

An Azure OpenAI-powered agent inside Halo receives a ticket about a server alert.

Agent calls an external MCP server (e.g. your monitoring tool) to pull live server status.

Agent uses that context to auto-categorise, prioritise, and add a diagnostic note to the ticket.

Resolution starts before a human agent has even looked at it.

Orchestration

One Prompt, Many Tools

The reason MCP matters isn't the single ticket it raises. It's that one instruction can read from one system, decide, and act in another — and that the connection between them is a standard rather than a script you maintain.

51%

of sales leaders using AI say disconnected systems are slowing their AI initiatives down — and 74% of sales professionals are cleaning data across siloed systems instead of selling.

Source: Salesforce, State of Sales, 2026 (vendor research). View the report

Working today

Several tool calls in one instruction

A single prompt can chain calls in succession — pull figures out of the reporting suite, then take action on what it found, without a second instruction from the user. Halo's own walkthrough runs this on OpenAI as the model.

“Show me every ticket breaching SLA on my team, add a chase note to each, and reassign the two oldest to me.”

Read, decide, act — one thread, no interface.

Halo's stated direction

The same prompt, across the whole stack

Halo has said it is actively working with other vendors on cross-platform orchestration — Halo calling out to their MCP servers and back. Halo positions this as in development rather than shipped, and we'll describe it the same way. What it looks like:

Ask for the assets with the most tickets raised against them in the last three months.

Halo returns the list — say, the worst twenty.

Raise a ticket against each one, in the same conversation.

Call out to the monitoring or RMM tool through its own MCP server and run a health check on each device.

Post the results back onto the matching Halo tickets.

Why a protocol beats a script

The alternative is hard-coding against each vendor's API. That works until an endpoint changes, and then someone has to find and fix it. MCP is deliberately looser: you describe the tools once and let the model work out how to use them, so the integration bends instead of breaking when the systems either side of it move.

Not just tickets

It Reaches Further Than the Service Desk

Halo is one platform on one database. Tickets, contracts, invoices, opportunities and customer conversations sit in the same record — so an assistant connected to Halo isn't limited to IT. The commercial side of the business is on the other end of the same endpoint.

One prompt, from Halo's own walkthrough

“Generate me a report of potential upselling opportunities among my customers — and generate an opportunity ticket for each one.”

Upsell report generated · opportunities created in the pipeline

That is one instruction doing two jobs — analysis, then creation. The AI reads the customer record, works out where there's room to grow, and writes the opportunities back into the pipeline. Deals a team would otherwise have found manually, or not at all.

It works because the assistant isn't looking at a CRM bolted onto a service desk. Support history, contract value, billing and open deals are attributes of the same customer, in the same database.

The sales story in full lives on Pipeline & opportunities →

Halo CRM →

Pipeline, quoting, campaigns and account management — a real sales CRM on the same platform and the same endpoint.

Halo CSM →

Customer service across every channel, with virtual agents, self-service and SLAs that an assistant can query and act on.

Reporting & dashboards →

The reports you expose to MCP are built here — the query builder, dashboards, scheduling and Power BI.

Why Allied ESM

We only do Halo. So we go deeper than anyone else.

We're not a generalist partner fitting MCP into a broader practice. Halo is all we do — which means we've configured this across multiple customers, know the data model inside out, and don't waste time on a learning curve.

Halo-only focus

We work exclusively with Halo. MCP isn't something we've bolted onto a generalist practice — it's a capability we've built and delivered for real customers, on real Halo instances.

We know the Halo data model

Configuring Tool Collections well means understanding how Halo's data is structured. We do — so custom tools, permission scopes, and AI connections are set up correctly first time.

Start fast, go further

No months-long discovery. We scope what's possible on your specific Halo version, connect your AI tool of choice, and get you to a working integration quickly — then build from there.

12+

Built-in tools available to any connected AI assistant, out of the box

Lines of custom code needed to connect an AI assistant to Halo

Any

AI tool that supports the open MCP standard can connect to Halo

v2.232+

Available now on current hosted Halo instances. Allied ESM can confirm your version.

FAQ

Common Questions

What is MCP in the context of Halo?

MCP stands for Model Context Protocol — an open standard that lets AI assistants interact with external systems. Halo exposes an MCP endpoint that any compatible AI tool can connect to, allowing it to discover and call Halo functions: raising tickets, searching the KB, logging time, and more. The AI acts on your behalf inside Halo, through conversation.

Which AI assistants work with Halo MCP?

Any AI assistant that supports the MCP standard can connect — including ChatGPT, Anthropic's Claude, and Microsoft Copilot Studio. The standard is open and adoption is growing rapidly across the AI ecosystem. If your organisation already uses an AI tool, there's a good chance it can connect to Halo MCP today.

Is Halo MCP secure?

Yes. MCP uses OAuth 2.0 for authentication and you can create separate Tool Collections with different permission scopes. A user-facing endpoint might only allow raising tickets and searching the KB, while an agent-facing endpoint has broader capabilities. Each collection has its own unique URL and access controls, so you're not exposing everything to everyone.

Do we need to build anything custom?

No custom development is required. Halo MCP is a built-in feature you enable in Configuration. Allied ESM can scope and configure it for you — enabling the endpoint, building the Tool Collections, connecting your AI assistant and testing the integration. That configuration work is scoped and quoted as part of your project.

How does this differ from the Halo Virtual Agent?

The Halo Virtual Agent is a chat interface built inside the Halo self-service portal, powered by AI. MCP is an open protocol endpoint that lets external AI assistants — tools your team already uses — call Halo directly. They're complementary: the Virtual Agent handles in-portal conversations, MCP brings Halo capabilities into whatever AI tool your organisation prefers. Both can be configured and deployed by Allied ESM.

Can the AI see all of our Halo data?

No — and that's by design. Access is scoped three ways. The AI can only call the functions on the tool list you configure. It can only read the data behind the specific reports you expose to it from Halo's reporting suite, not the wider database. And each Tool Collection has its own endpoint, OAuth application and audience, so an end-user endpoint and an agent endpoint never share permissions. One caveat worth knowing: the in-Halo chat terminal currently inherits permissions from the account the integration authenticates as rather than the individual agent, which is why Halo recommends keeping that particular deployment to admin agents for now.

Does Halo MCP only work for IT tickets?

No. Halo is a single platform on a single database, so the same endpoint reaches sales, customer service and managed-services data as well as the service desk. In Halo's own walkthrough, one prompt produces a report of upsell opportunities across the customer base and creates an opportunity record for each — a commercial task, run through the same MCP connection an agent uses to raise a ticket. See Pipeline & opportunities for the sales side and Halo CSM for customer service.

Ready to Connect Your AI to Halo?

Whether you want to start with a single use case or design the full AI stack, Allied ESM can scope and deliver it. Get in touch and we'll walk you through what's possible on your specific Halo version.