Research on request
Compare, summarise, cite. The result is filed, not lost to a closed tab.
“What changed in the EU AI Act this month? Two paragraphs, with sources.”
private agents, set up & managed
Your own private AI agent, set up for you.
Message it like you’d message anyone. It researches, remembers, keeps your files in order and runs your recurring jobs — while Hervo quietly handles the hosting, updates and recovery.
£250 one-time setup£99/month, kept runninglive within two working days
Founding customers only, for now — no checkout, just a call. Each agent is set up by hand.
Albertyour agent · managed by Hervo
on dutyTue · 07:30
Scheduled · weekdays 07:30
Morning brief: two replies waiting on the Meridian thread, your Kestrel invoice is due Friday, and three new pieces on the freelance tax changes — summaries filed in Research / Tax.
09:14Morning brief: two replies waiting on the Meridian thread, your Kestrel invoice is due Friday, and three new pieces on the freelance tax changes — summaries filed in Research / Tax.
Draft the renewal email for Dana — same terms, friendlier open.
Memory · March notes + pricing sheet
Drafted. I kept the twelve-month rate from your pricing sheet and mentioned the workshop she liked in March. It’s in your drafts — say the word and I’ll tidy the subject line too.
Drafted. I kept the twelve-month rate from your pricing sheet and mentioned the workshop she liked in March. It’s in your drafts — say the word and I’ll tidy the subject line too.
albert · private instance — worklog
07:30routinemorning-brief done · 3 sources
07:31filesfiled 3 summaries → /research/tax
09:14memoryrecall “dana renewal” · 2 notes, 1 file
09:15draftrenewal-email.md → saved to drafts
02:00backupmemory + config + skills ✓ nightly
What changes
You already work with AI all day. What’s missing is continuity.
If ChatGPT or Claude is open in a tab right now, you don’t need convincing that the intelligence is real. An agent of your own fixes everything around it.
Your chats todayEvery chat starts from zero.
With your own agentYours already knows your clients, projects and preferences.
Your chats todayYou paste the same context in, again and again.
With your own agentYour documents and notes are part of its memory.
Your chats todayGood research vanishes when the tab closes.
With your own agentFindings are filed where you'll find them next month.
Your chats todayNothing happens unless you're at the keyboard.
With your own agentBriefings and checks run on the schedule you set.
Your chats todayAnother app, another tab, another login.
With your own agentIt's a chat in the messenger you already use.
Concretely
Things you can actually hand over.
No vague superpowers. These are the jobs your agent is configured for on day one — each one triggered by a plain message.
Remember what matters
Clients, prices, preferences, past decisions — recalled exactly when they're relevant.
“What did we quote Meridian last year, and did they push back?”
Keep files in order
Naming, filing and finding documents on request, in folders you both understand.
“File these receipts under Q3 expenses, then total them.”
Run recurring routines
Morning briefs, weekly reviews, deadline sweeps. Three are set up on day one.
Every Mon 08:00 — “Week ahead, open invoices, anything overdue.”
Draft and prepare
Emails, proposals, agendas — in your tone, built from your own material.
“Draft a follow-up from today's call notes. Keep it short.”
Watch and remind
Dates, renewals, follow-ups — surfaced before they turn urgent.
“Nudge me three days before any invoice falls due.”
It’s an agent, not an employee — it works from the memory, tools and routines you set up together. See the boundaries.
How setup works
One call, two working days, then it’s yours.
You never touch a server, an API dashboard or a config file. That part is the product.
Step 01
Book the call
Forty-five minutes. You choose a name for your agent, the messaging app you'll talk to it in, and up to two integrations. A short questionnaire gathers context and credentials afterwards.
Step 02
We build your agent
A dedicated private instance: model and provider configured, your profile and initial memory loaded, research and file tools connected, three scheduled routines wired in and tested.
Step 03
Say hello
Your agent introduces itself in your messenger, we walk you through what it knows and when it will speak up — and we stay close for a full week of launch support.
Pricing
Two numbers. No tiers, no seats, no asterisk farm.
One payment to set your agent up properly, one subscription so it keeps working without becoming your technical project.
Setup · one-time
£250 once
Everything needed to go from “interesting idea” to a named agent answering you by name.
- A dedicated private agent instance — yours alone
- Your choice of one supported messaging interface
- Model and provider configured to suit your work
- Personal profile and initial memory loaded
- Web research and file tools connected
- Up to two supported integrations
- Three scheduled routines that earn their keep
- A 45-minute onboarding call
- One week of launch support
→ delivery target: two working days after we receive your questionnaire and credentials
Book a setup callManaged · monthly
£99 / month
The part you’re really buying: your agent still works next month, and the month after that.
- Hosting and monitoring of your instance
- Updates as the underlying tools evolve
- Backups of configuration, memory and skills
- Recovery if the instance ever breaks
- Model and provider configuration kept current
- Up to 30 minutes of support or small changes monthly
- One new or adjusted scheduled routine each month
→ cancel monthly; your agent’s memory and configuration stay backed up throughout
Third-party model/API usage is billed separately by the provider — bring your own keys or we’ll set billing up in your name. Custom integrations and bespoke workflows are scoped and quoted separately.
What Hervo isn’t
You’re paying for exactly one thing: a private agent that keeps working without becoming your technical project.
- Not a replacement for your assistant, and it won’t run your business by itself.
- Not unlimited usage, and not around-the-clock support.
- No promised “hours saved per week” figures — you’ll know within a month whether it earns its keep.
- No enterprise compliance certifications or SLAs. This is a personal service for individuals; if you need formal guarantees, we’re not there yet.
Ownership
Private means yours. Plainly.
i.One instance, one owner.
Your agent runs on its own dedicated instance — not a shared bot with your name stencilled on. Its memory and files live there, for your work only.
ii.Your keys, your models.
You choose the model provider, and you can bring your own API keys. If a better model appears next quarter, we reconfigure — you don't re-buy anything.
iii.It acts on rails.
Your agent replies when you message it and runs the routines you approved. Integrations reach only what you connected. It has no authority you didn't give it.
iv.Backed up and recoverable.
Configuration, memory and skills are backed up as part of the monthly plan. If the instance breaks, we rebuild it from backup — that's our job, not yours.
We won’t dress this page in compliance badges we haven’t earned. Model providers’ own terms apply to API usage, and if you work in a regulated environment that needs formal guarantees, ask us on the call — we’d rather say “not yet” than pretend.
Questions
Asked and answered.
Couldn't I build this myself with open-source tools?
Genuinely, yes — the underlying pieces are public, and if you enjoy hosting, securing and maintaining software, you should try it. Hervo exists for people who want the agent without the second job. You're paying for setup done well and for someone else being on the hook when it breaks.
What do the AI model costs look like?
Model/API usage is billed by the provider and isn't included in the £99 — most personal workloads land in the tens of pounds a month, but it depends on how hard you run it. You can bring your own API keys or we'll set billing up in your name, and we configure sensible defaults so costs stay boring.
Which messaging apps can I use?
You pick one supported messaging interface during setup — we keep a short list of familiar apps and confirm the best fit on the call. The point is that talking to your agent should feel like texting a person you already message, not opening yet another tool.
What integrations are included?
Up to two supported integrations are part of setup — think calendar, email, notes or file storage. Anything custom or unusual is scoped and quoted separately, honestly, before any work starts.
Will it act on its own?
Only in the ways you've approved: it runs the scheduled routines you set up (a morning brief, a weekly review, a deadline sweep) and otherwise responds when you message it. It doesn't freelance, spend money or contact people on its own authority.
What support is included after launch?
One week of close launch support after handover, then up to 30 minutes of support or small changes each month, plus one new or adjusted routine monthly. Bigger changes are quoted separately. If your instance breaks, recovery is on us — that's the core of the monthly plan.
What happens if I cancel?
The monthly plan is month to month. Your agent's configuration and memory are backed up throughout, so nothing about your data is held hostage — we'll wind the instance down properly and hand you the latest backup.