Included in Halon Engage and Halon Protect 26.3 release
We’re excited to launch our latest release for Halon Engage and Halon Protect 26.3. We’re introducing the first version of Halon’s Model Context Protocol (MCP) integration, bringing AI-assisted capabilities to your email operations. It complements the specialized intelligence in components like Delivery Guru, which acts in real time as email is processed, by enabling AI agents to analyze your system and operational data in greater depth outside the email flow. Your team gets greater insights, helping you investigate issues more efficiently, spot emerging delivery and abuse patterns earlier, and take more informed actions.
The integration connects the MCP-compatible AI agents your team already uses directly to your Halon environment. Halon’s comprehensive suite of APIs gives your agent a rich set of data sources to call. You can ask questions in plain language and get immediate, complete responses, schedule AI tasks, set up intelligent monitoring, create bespoke reports and dashboards, and much more. Instead of manually checking each tenant or destination, your team can focus its attention on the issues that need investigation. You just ask!
Halon’s MCP integration advances Dynamic and Secure Email Operations™ through faster issue resolution and intelligent, proactive monitoring. Teams can investigate issues through plain-language questions, reducing reliance on specialists for routine analysis. Automate routine checks, bring operational data together, and identify emerging delivery and abuse issues earlier, so your team spends less time investigating and can make informed decisions sooner.
This initial release includes a library of annotated calls tested against real-world email operations tasks, read-only access to begin with, starter skills for Halon Engage and Halon Protect, and a straightforward setup. The MCP is part of the web administration (halon-web package), which means there’s no additional installation needed, and no separate service to host or wrappers to maintain.
What MCP adds to your email operations
Working with a traditional API typically means writing code, because APIs are primarily built for applications. MCP is like an API built for agents: you describe what you want in plain language, and the agent works out which calls to make.
To be genuinely useful, though, an agent needs three things:
- Understanding: traditional API documentation is not written for AI agents; they work best with clearer descriptions of what each call does and when to use it
- Know-how: a skill that explains how to carry out email tasks, filling gaps in the AI model’s training
- Context: access to the other data a task depends on, like your logs or threat intelligence
Halon’s MCP integration is built to deliver the first, give you a head start on the second, and work hand in hand with the third.
Let's take a closer look at what's included in the Halon Engage and Halon Protect releases.
An annotated call library
Halon has built a library of annotated calls and responses across its composable email infrastructure and tooling. Each call carries a plain-language description of what it does in Halon and where to find the data you need, helping your agent identify the relevant calls for the task faster.
We tested these descriptions against real email operations tasks, and we’re continuing to refine them. Structured descriptions also reduce ambiguity, making it easier for agents to select the right calls and deliver more consistent, accurate answers.
The first release covers more than 20 tools, from live queues, policies, suspensions, and connections to stored delivery history.
Read-only by design, at first
The first version is read-only on purpose. Your agent can read your data and report on it, but it can’t do things like change your configuration or take action on queued messages. That means you can put AI to work on your email operations without giving agents the ability to modify your production setup.
Your agent signs in to Halon through OAuth, the standard way MCP handles access. It gets the same permissions as the Halon user account you connect it to. For tighter control, consider giving the agent its own account and appropriate data masking options, with access limited to only what it needs.
The limit then applies wherever that account signs in, not just to MCP. Your agent also sees only the delivery history fields you have chosen to expose in your Halon configuration.
Starter skills
Your agent gets a head start from day one. Halon provides two initial starter skills, one for Halon Engage and one for Halon Protect, each covering common email operations tasks for that product. You can adapt them to your own workflows and build on them over time.
Get the full picture of your email operations
Halon’s MCP integration works alongside the MCP integrations for your other email tools. Connect your logging platform or SIEM, and your agent can query them all in one conversation. Ask how today compares to last week, and you get live operational data from Halon and history from your logs, side by side. That means less switching between tools and manually piecing together what happened, helping your team get to the cause and decide how to respond.
Halon’s own delivery history is part of the integration too. Your agent can search past deliveries, follow one message from start to finish, or run ready-made reports on bounces, deferrals, top errors, and delivery times.
Simple to set up
The Halon MCP server comes with the 26.3 release. Once you’ve upgraded, switch on the server in your configuration, connect your agent through OAuth, and add the starter skill for your solution. There’s no separate service to host and no wrappers to maintain.
Resources: [Installation guide, Upgrade guide, Configuration reference, and changelog]
MCP with Halon Engage 26.3 - Investigate delivery issues faster
High-volume senders often solve delivery problems by correlating data to spot the pattern behind them. When your email drives revenue, every minute spent piecing the picture together costs you.
Halon’s MCP integration makes querying that data, and getting answers from it, much easier. Just ask about your delivery in plain language. Here are some examples we’ve explored:
You want a monitoring agent to spot early signs of delivery or abuse issues
Q: “Notify me when failure rates or queue sizes spike for any tenant or destination, so I can spot a compromised tenant or a destination where several tenants are having delivery problems.”
A: A scheduled check compares failure rates, deferrals, and queue sizes for each tenant and each destination, and flags the ones that stand out. That way, a compromised tenant, or a destination where messages from several tenants are having delivery problems, shows up early.
You’ve noticed mail to a mailbox provider is backing up
Q: “Why is mail to {mailbox provider} backing up?”
A: Whether it’s your own rate limit, the provider pushing back or a suspension, with the evidence behind it.
You want a bespoke, live-updating dashboard for your deliverability team, tailored to your business and team priorities
Q: “Can you build me an interactive dashboard of my current email status for today, warmups, deferrals, and blocks so I can review with my team? I would like to auto-refresh every 10 minutes.”
A: A dashboard built from live queue, suspension and warm-up data alongside delivery history, laid out the way your team works.
Your customer asks why their mail is slow
Q: “What rate policy applies to the tenant {tenant name}?”
A: The tenant’s policy and queue quota, so you can see where the issue lies before you reply.
You’re warming up new IPs
Q: “Which IPs are currently in warm-up, and where are we seeing deferrals?”
A: Warm-up progress and deferrals side by side, so you can decide whether to slow down.
You’re planning a change
Q: “Are all nodes on the same version?”
A: The software version on each node. Ask before you start and again when you’re done. On our demo cluster, it spotted one host running a different version from the rest.
You’re onboarding a new sending domain
Q: “Are SPF, DKIM and DMARC published for {domain}?”
A: Which records are published in DNS, so a missing one shows up before the first send.
You want to know why mail bounced
Q: “How many messages bounced at Gmail today, and why?”
A: Bounce counts and the top reasons, straight from Halon’s delivery history. In testing, the agent found 4,775 Gmail bounces in 24 hours and ranked the causes, led by invalid recipient addresses.
Support asks where a customer’s message went
Q: “Is there mail for {recipient} from tenant {tenant name} in the queue, and why hasn’t it been delivered?”
A: The message’s current state, the last delivery attempt and the response from the receiving server, in plain language, ready to pass on.
Every answer cites its evidence, and if something’s missing, it tells you. Patterns become easier to spot, so you can craft the right response sooner.
Watch the demo: Halon’s MCP integration with Halon Engage
Investigate delivery issues in plain language with Halon Engage 26.3.
MCP with Halon Protect 26.3 - Assess email threats faster
Halon Protect already keeps threats out of your users’ inboxes. Halon’s MCP integration makes it easier to see how. Ask about your policies, your configuration or recent inbound mail in plain language, and get a clear answer with the evidence behind it. Here are some examples we’ve explored:
You want to check your setup is sound
Q: “Are there any errors or inconsistencies in my configuration?”
A: Any issues the agent finds, along with their location in your configuration.
You suspect a spam wave hit overnight
Q: “Was there a spam wave last night, and who did it reach?”
A: What arrived, when and who was affected. In testing, the agent found a burst of a few hundred messages, confirmed most were blocked, and flagged the one user who may have received one.
You need to know how inbound mail for a domain is being filtered
Q: “How is inbound mail for {domain} being filtered, and what's being caught?”
A: Each message’s verdict and final action from Halon Protect’s delivery history, across spam and virus scanning, Spamhaus, SPF and DMARC checks.
You’re investigating a spam campaign in depth
Q: “Give me the delivery profile of this campaign and the users it affected.”
A: The campaign’s delivery profile and affected users from Halon Protect. Connect other MCP integrations, such as threat intelligence or IP and geolocation data, and your agent can bring that context into the same investigation.
These examples show how Halon enables Dynamic and Secure Email Operations™ in practice: scheduled checks help teams spot emerging issues, reusable skills reduce repetitive investigation, and connected data sources bring delivery and security insights into one conversation. Your team gains more time to act on what matters, with even greater insight and control.
What’s next for Halon’s MCP integration
We’ve released a read-only MCP server that agents can connect to. This is just the start. Here’s what we’re working on to make it even better:
- Continuing to expand the call library and testing it against real email operations tasks to fine-tune its descriptions.
- Learning from teams using the integration in their daily operations, to understand what’s working and where it needs to improve.
- Using that feedback to improve tool discovery and annotations, and to develop more advanced agent skills.
No one knows your workflows better than you. When the barrier is a plain-language prompt instead of code, you can get creative and try things you never would have scripted before. We’d love to hear what you’re building and where it could improve.
To see a live demo of Halon’s MCP integration, schedule a call with one of our email experts.