
StackJack is a managed MCP (Model Context Protocol) platform built for MSPs and IT teams. It connects the AI your techs already use (Claude, ChatGPT, Microsoft 365 Copilot, Copilot Studio, Cursor, n8n, or anything else that speaks MCP) to the systems that run the business, through one endpoint, with the controls you need before you let AI anywhere near production data. HaloPSA was one of StackJack's two launch connectors and it is still the deepest: 1,551+ tools covering tickets, actions and SLAs, clients, sites and users, assets, contracts and recurring invoices, CSP licensing, quotes, sales and purchase orders, projects, timesheets, the knowledge base, reports, custom tables, and the configuration area itself.
Connecting Halo to StackJack gives you three things.
The proxy. Your AI never holds your Halo credentials. Every tool call passes through StackJack's MCP proxy, and the credentials live in Azure Key Vault, not in a database and not on a tech's laptop. With Authorization Code sign-in, each technician connects with their own Halo agent, so every change the AI makes lands in Halo's audit trail under the person who asked for it, with that agent's Halo roles still enforced. You decide exactly which tools each person, role and AI client can see, and the Free Tier is read-only, so nothing writes to Halo until you decide it should. Requests are paced to 300 per 5 minutes per organization with automatic backoff, so an overeager AI can't hammer your instance, and the proxy logs call metadata for the audit trail, not payloads.
Automations. StackJack hosts Claude-powered agents that work inside Halo on their own, started on demand, on a schedule, or by a webhook. Point a Halo webhook at an automation and a new ticket gets triaged, enriched from your RMM and documentation, categorized and routed before a human opens it. Schedule one to check contracts against recurring invoices every Monday, chase stale tickets, or reconcile CSP seat counts against Halo licence records. Every new automation starts in dry-run mode, where write calls are recorded as "would have called" instead of executed, and it only writes to Halo after you explicitly promote it to production. Runtime and spend caps, an explicit tool allow-list, run transcripts, version history and rollback come standard.
AI Governance. Once your team is using AI against Halo, somebody has to answer "what did the AI see, and what did it send?" StackJack AI Governance checks tool traffic against your data-loss rules as each call happens, in both directions: what goes into Halo and what comes back out. Depending on the rule, StackJack blocks the call, masks the matched value before the AI ever sees it, or makes the technician give a reason before continuing. The same console reads AI usage from Microsoft 365 Copilot, Azure OpenAI and Foundry, Island, Cloudflare AI Gateway and DefensX, so you get one view of AI use across the business, not just the Halo slice.
Halo stays the source of truth. StackJack makes it safe to let AI read it, write to it, and act on it.
Halo is one of over 114+ connectors on StackJack (NinjaOne, CIPP, IT Glue, Hudu, Microsoft Graph and more), so a single question or automation can pull from Halo and the rest of your stack in the same run. Pricing is per connector, never per seat, and every connector has a Free Tier.
- 1,551+ (and growing all the time) HaloPSA tools across tickets, clients, assets, contracts, billing, sales, projects, timesheets, knowledge base, reporting and configuration
- Free Tier with every Halo read tool; Pro and Business add write tools and higher monthly call volume
- Per-technician sign-in (Authorization Code) so AI actions are attributed to the real agent in Halo's audit trail, or Client Credentials with a dedicated application identity
- Halo credentials held in Azure Key Vault, with automatic token refresh and background validation
- Custom tool roles controlling exactly which Halo tools each technician, team and AI client can use
- API Permissions planner mapping every tool to the exact Halo API permission it needs, for least-privilege API applications
- Request pacing (300 per 5 minutes per organization, backoff on 429s) to protect your Halo instance
- Response shaping with a HaloPSA preset that trims bulky schema metadata so the AI spends its context on real work
- Multiple Halo connections on one connector, choosing the instance per call or pinning a client to one
- One Streamable HTTP endpoint for Claude, ChatGPT, Microsoft 365 Copilot, Copilot Studio, Cursor, VS Code and n8n, plus compact and per-connector endpoints for clients that cap tool counts
- Microsoft CSP and M365 subscription questions answered from Halo's own licence records
- Automations triggered on demand, on a schedule, or by Halo webhooks
- Build automations from templates, a guided wizard, the advanced builder, or straight from your AI assistant
- Dry-run by default, supervised test runs with approval of destructive calls, and explicit promotion to production
- Runtime and spend caps, outbound network limits, run history with transcripts, version history and rollback, and failure notifications
- Bring your own Anthropic key on eligible plans
- AI Governance data-loss rules on Halo tool traffic: alerts on Pro; block, mask, or warn-with-reason on Business and above
- Optional fail-closed mode that refuses any call StackJack cannot finish scanning
- AI usage visibility across Microsoft 365 Copilot, Azure OpenAI and Foundry, Island, Cloudflare AI Gateway and DefensX
- Searchable, exportable audit log of every tool call
- US and EU data regions with self-service migration
- SOC 2 Type II Audited (Security, Availability, Confidentiality), report available at trust.stackjack.io