Is our data used to train models?

No. Tenant data is isolated per-tenant and never leaves your deployment to train a shared model. On Core, inference runs against Claude via the Anthropic API under a zero-retention agreement. On Sovereign, you can bring your own keys or host the LLM yourself — your prompts and completions stay inside your environment.

What happens to our tickets if we leave?

You can export every ticket, device, runbook, audit log entry, and chat transcript as JSON + CSV from the admin console at any time, with or without staff help. The pilot contract explicitly grants you a full data export on exit — no migration fee, no lock-in clause.

How do we know the AI won't break production?

Every AI-proposed action carries a risk level (LOW / MEDIUM / HIGH / CRITICAL) and a confidence score. Only LOW-risk, high-confidence actions auto-execute; anything else is held in an approval queue with a diff and a predicted blast radius. Every action is logged with the prompt, the tool calls, the approver, and the outcome — you can replay or roll back any change from the audit trail.

What about regulated workloads?

The Sovereign tier runs in a single-tenant UK or EU deployment (no shared infra, no data crossing the Atlantic). It ships with SOC 2 / ISO 27001 evidence packs, Cyber Essentials Plus walkthroughs, and a dedicated customer engineer for change-control sign-off. If your regulator requires an on-prem LLM, that's the Phase 2 roadmap.

How does pricing scale?

£12 per managed endpoint per month on Core, flat — no per-module upsell, no seat-count games, no ticket-volume surcharges. Sovereign starts at £38/endpoint/mo and includes the dedicated engineer and evidence packs. Pilot is free for 60 days on up to 50 endpoints.

What if an endpoint is offline?

Agents run locally in Rust with a <15 MB footprint and an SQLite offline buffer. Telemetry, heartbeats, and command results queue locally and drain when the device reconnects. A device going offline raises an alert, not a silent data gap.

LIVE · 1,247 tickets resolved today

The MSP that
fixes itself.

An AI-first managed services platform that resolves 4 out of 5 tickets before a human ever sees them — so your team stops triaging passwords and starts running a business.

Book a demo Watch 90-second tour

✓ Self-hosted option
✓ Cyber Essentials+
✓ 30-day pilot, no lock-in

TICKET #4821Can't connect to shared drive

RESOLVED

09:14:02UserShared drive not mounting after login. Windows 11, remote.
09:14:03Infra agentPulled device telemetry. DNS cache miss + stale Kerberos ticket. High-confidence match to runbook #203.
09:14:04IT Support agentProposed: flush DNS, renew ticket, remount X:\. Risk: LOW. Auto-approved.
09:14:09EndpointExecuted in 4.1s. Drive mounted. Verifying…
09:14:15Fortuna PaceSent user Teams message. Closed ticket. Updated runbook confidence: 97% → 98%.

Resolution

13s

Engineer cost

£0.00

Human touches

Pilot cohort · Q1 20266 months · 5 MSPs · 3,200+ endpoints

Five MSPs running Fortuna Pace today, across 3,200 endpoints. Here's what that looks like.

Axiom
software · 48 seats
Bastion
infrastructure · 120 seats
Cipher
finance · 85 seats
Keeper
healthcare · 220 seats
Lumina
creative · 32 seats

We went from 6 Tier-1 engineers to 2. The other 4 moved up to security and automation work — the stuff we actually pay them for.

Priya Ahuja

Ops Director, Axiom

81%

Autonomous resolution (pilot avg)

Lower bound across 5 tenants. Range 73–88%.

4.3×

Tickets per engineer / day

vs. pre-rollout baseline on the same fleet.

18s

Median L1 resolution

Known issues. P95 under 3 minutes.

ISO 27001 (in progress)
SOC 2 Type I (Q3 2026)
Cyber Essentials Plus
UK-sovereign hosting
Methodology: /pilot-report

How it works

From a ticket to a fix in six deterministic steps.

Every resolved ticket walks the same path. Risk and confidence gate execution, and the audit log makes the whole thing replayable.

01 / 06
Telemetry
Rust agents stream device health, logs, and inventory over a lightweight channel — every signal is timestamped and tenant-scoped.
02 / 06
Orchestrator
A CTO persona classifies the intent, checks permissions, and picks the right specialist agent for the job.
03 / 06
Specialist agent
IT support, security, infrastructure, automation, or business ops — each agent has its own tools, runbooks, and risk ceiling.
04 / 06
RAG + tools
The agent grounds its plan in your runbooks and prior ticket resolutions, then calls approved tools: script runners, integrations, queries.
05 / 06
Endpoint
LOW-risk, high-confidence actions auto-execute on the device. MEDIUM/HIGH hold for approval with a predicted blast radius.
06 / 06
Close
The platform verifies the fix, updates the runbook confidence score, notifies the user, and writes the audit trail.

Explore AI platform

6 tools90% confidence gate<15min auto-resolve

The console

One pane. Every tenant.
No sixteen tabs of ConnectWise.

Real-time fleet health, AI-proposed actions you approve or reject with a keystroke, and a complete audit trail for every command the platform ever runs. Built for an engineer who has 40 clients open at once.

⌘K
Command palette
Jump to any device, ticket, or tenant in under 200ms.
●
Real-time fleet
WebSocket stream; dashboards never go stale.
✓
Approval gating
High-risk actions are held for review with diffs and predicted impact.
↻
Full audit log
Every AI action, who approved it, what changed. Compliance-ready.

console.fortunapace.com / halcyon-labs

Healthy

+2 24h

Needs attention

all self-healing

AI resolving

avg 23s

Awaiting approval

Priya · 2m

Resolution mix · last 24h84.2% autonomous

AWAITING APPROVAL

Restart SQLSERVER on srv-fin-01.

Risk: MEDIUM · Confidence: 94% · Affects 12 users · Predicted downtime 45s

Active tickets12 open · 3 human · 9 AI

#4821Drive mount · FinanceAI · 13sresolved

#4820VPN drop · RemoteAI · in progressworking

#4819Printer offline · HQQueued · Priyahuman

#4818Outlook calendar syncAI · awaiting approvalhold

Pricing · simple · per endpoint

One number, per endpoint, per month.
No tiered feature gates. No “contact sales” smoke.

MSPs are tired of ConnectWise-style per-module pricing. We charge one price per managed endpoint. AI resolution, monitoring, tickets, billing, compliance — all included. Bring your own keys, bring your own LLM in Sovereign.

START HERE

Pilot

Free

60 days · up to 50 endpoints

Full platform access
Managed rollout with our team
Weekly resolution report
No credit card, no commitment
Export everything on exit

Start pilot

We benchmark against the incumbents, publicly.

If you've used ConnectWise or NinjaOne for a decade, the switching cost is real. We're transparent about where we win, where we're even, and where we're still catching up. Updated monthly.

Capability	★ Fortuna Pace	ConnectWise	NinjaOne	Syncro	Atera
AI-first architecture	Built-in	Add-on (Sidekick)	Add-on (Ninja IQ)	Bolt-on LLM	AI Copilot beta
Autonomous resolution	80%+ target, multi-agent	Chat-only suggestions	Suggestions, no exec	Chat summarisation	Triage only
Endpoint agent	Rust · <15MB	Legacy .NET	Proprietary	Split Win/Mac	Proprietary
Pricing model	Per endpoint, all-in	Per module	Per endpoint +add-ons	Per tech	Per tech
Self-host / sovereign	Sovereign tier	—	—	—	—
Compliance automation	Cyber Essentials+, ISO 27001 packs	Separate product	Third-party	Third-party	Third-party
Approval-gated AI actions	Every action, diffs + rollback	N/A	Manual scripts	Manual scripts	Limited
Time to first value	Day 1 (managed pilot)	4–8 weeks	2–3 weeks	1–2 weeks	Same day

Last verified: Apr 2026 · Sources: vendor docs, pilot customers migrating inSee full methodology →

Frequently asked

The questions MSPs ask before switching.

Is our data used to train models?
No. Tenant data is isolated per-tenant and never leaves your deployment to train a shared model. On Core, inference runs against Claude via the Anthropic API under a zero-retention agreement. On Sovereign, you can bring your own keys or host the LLM yourself — your prompts and completions stay inside your environment.
What happens to our tickets if we leave?
You can export every ticket, device, runbook, audit log entry, and chat transcript as JSON + CSV from the admin console at any time, with or without staff help. The pilot contract explicitly grants you a full data export on exit — no migration fee, no lock-in clause.
How do we know the AI won't break production?
Every AI-proposed action carries a risk level (LOW / MEDIUM / HIGH / CRITICAL) and a confidence score. Only LOW-risk, high-confidence actions auto-execute; anything else is held in an approval queue with a diff and a predicted blast radius. Every action is logged with the prompt, the tool calls, the approver, and the outcome — you can replay or roll back any change from the audit trail.
What about regulated workloads?
The Sovereign tier runs in a single-tenant UK or EU deployment (no shared infra, no data crossing the Atlantic). It ships with SOC 2 / ISO 27001 evidence packs, Cyber Essentials Plus walkthroughs, and a dedicated customer engineer for change-control sign-off. If your regulator requires an on-prem LLM, that's the Phase 2 roadmap.
How does pricing scale?
£12 per managed endpoint per month on Core, flat — no per-module upsell, no seat-count games, no ticket-volume surcharges. Sovereign starts at £38/endpoint/mo and includes the dedicated engineer and evidence packs. Pilot is free for 60 days on up to 50 endpoints.
What if an endpoint is offline?
Agents run locally in Rust with a <15 MB footprint and an SQLite offline buffer. Telemetry, heartbeats, and command results queue locally and drain when the device reconnects. A device going offline raises an alert, not a silent data gap.

See it work on your fleet.

Twenty minutes, one of your tenants, the same console every pilot customer uses. You bring a recent ticket; we'll run it through the agents live.

Book a 20-minute demo

No sales deck. No slides. A ticket and a clock.