An open-source AI agent on your own server.
We install the OutGrow AI Employee, connect it to your tools and keep it running safely. A person approves anything the agent is unsure of. Part of a partnership, or set up on its own.
~2 wks
To go live
$4,500
One-time setup
Custom skills
Installed and looked after by OutGrow. That covers setup, security and ongoing care.
The admin that eats your week.
Time drain
Chasing emails, booking meetings and pulling reports eat the hours you meant to spend on the business.
Hiring pressure
More clients means more staff and more management, and payroll climbs faster than revenue.
Leads that slip
Follow-ups never happen and opportunities sit buried in an inbox.
Name the job and it works out how.
The AI Employee plans the work and acts through your tools rather than just replying.
Email triage
Reads, sorts and summarises your inbox, drafts replies and flags what needs you.
Calendar and scheduling
Books meetings, sorts out clashes and sends reminders in the calendar you already use.
Web research
Browses the web, fills in forms and pulls out data, then sends you a clean summary.
Files and documents
Writes reports, reviews documents and organises files so your data is structured and ready to use.
Background monitoring
Watches for errors, tracks deadlines and sends alerts, day and night.
Your apps
Connects to Gmail, CRMs, QuickBooks, GitHub, Todoist, Slack and calendars. We build new skills for the tools you rely on.
It runs on your server, under your rules.
Self-hosted, on infrastructure you choose
Your cloud account, a private server or a machine in your office. Files and history stay there, and any commercial AI models are called with your own API keys.
A person approves what the agent is unsure of
Confident actions go ahead. Uncertain ones go to a person with full context, and you set that line per workflow and action type.
Permissions on every outside action
Allowlists, sandboxing and approval limits mean the agent only touches what it's allowed to. It never gets unrestricted access.
Prompt injection protection
Security wrappers and input checks stop outsiders from hijacking the agent's instructions.
A log of every step
See what each agent did, which tools it called, how long each step took and what it cost. Any step can be replayed.
Memory, backups and recovery
Long tasks keep their state through restarts. Automated backups, rollback steps and alerts stop one fault taking your operations down.
Six open-source versions for different setups.
We deploy and manage any of them. Tell us your stack and constraints, and we'll recommend one.
Nanobot
"A lightweight personal assistant."
- About 4,000 lines of code, far smaller than the full framework
- MCP support with multi-channel integration
NanoClaw
"Small enough to understand in an afternoon."
- Agents run inside real Linux containers
- Supports agent swarms
IronClaw
"Built with security first."
- WASM sandbox isolates untrusted tools
- Suited to regulated environments
ZeroClaw
"Very low overhead."
- Runs on low-cost hardware with a tiny memory footprint
- Compiles to a single static binary
PicoClaw
"AI on almost anything that runs Linux."
- Runs on old Android phones and embedded boards
- Suited to IoT and edge devices
TinyClaw
"An always-on assistant for teams."
- Several agents across teams and channels
- Live terminal dashboard for monitoring
Five steps from first call to launch.
Days 1-3
Discovery and scoping
We look at how you work, find your five biggest time drains and agree what the agent takes on first.
Days 4-7
Installation and setup
We install the AI Employee on a secure server, connect your messaging apps and build custom skills for your business.
Days 8-10
Guardrails and security
Access controls, data boundaries, audit logs and prompt injection protection go in before launch.
Days 11-13
Testing and team training
We run the agent through real scenarios, fix edge cases and show your team how to hand it work.
Day 14 onwards
Go live and ongoing care
The agent starts work. We monitor it, tune it and add skills as your needs grow.
One setup fee, then an optional monthly plan.
Prices are in dollars. We can also quote in INR, AED or GBP.
One-time setup
$4,500 one-time fee
- Full private server or cloud deployment
- Custom integrations with up to 3 platforms
- Security guardrails and access controls
- 5 custom skills built for your business
- Team training and documentation
- 14 days of post-launch support
Monthly retainer
$860 / month, cancel anytime
- Agent monitoring and uptime checks
- Monthly performance tuning
- New skills as you need them
- Priority support with a 24-hour response
- Security patches and updates
- Usage and ROI reports
First year with the monthly plan: $14,820 in total.
Questions about the AI Employee.
How much does the OutGrow AI Employee cost?
A one-time setup fee of $4,500, then an optional monthly plan of $860 that you can cancel anytime. The first year with the monthly plan comes to $14,820. We can also quote in INR, AED or GBP.
How long does it take to go live?
About two weeks. Days 1 to 3 are discovery and scoping, days 4 to 7 installation and setup, days 8 to 10 guardrails and security, and days 11 to 13 testing and team training. The agent starts work from day 14.
Where does the AI Employee run?
On infrastructure you choose: your cloud account, a private server or a machine in your office. Files and history stay there, and any commercial AI models are called with your own API keys.
Which open-source framework is it built on?
We deploy and manage six: Nanobot (Python), NanoClaw (TypeScript), IronClaw (Rust), ZeroClaw (Rust), PicoClaw (Go) and TinyClaw (TypeScript). Tell us your stack and constraints, and we recommend one.
Which apps can it use?
Gmail, CRMs, QuickBooks, GitHub, Todoist, Slack and calendars, and we build new skills for the tools you rely on. Setup includes integrations with up to three platforms and five custom skills.
How do you stop it doing something it should not?
Confident actions go ahead and uncertain ones go to a person with full context; you set that line per workflow and action type. Allowlists, sandboxing and approval limits mean it only touches what it is allowed to, and every step is logged and can be replayed.
Try an agent on your own busywork.
Thirty minutes, free. We map the workflows worth handing over and show you the AI Employee running live.
Full name
Company
Work email
Phone
We start from how the work runs today, not from a tool.
Most businesses we meet have paid for an AI tool nobody uses. We start from your numbers, so the first build is chosen by payback.
Zain Khan and Anwar Sheikh
Founders, OutGrow
Outgrow
Intelligence
Studios.