Day 27 of 28 · OpenClaw Challenge
Sharing Your Agent
⏱ 7 min
📊 Intermediate
You've spent three weeks building a powerful, personalized AI agent. It handles your email, manages your calendar, runs workflows, and analyzes data.
Now imagine if your entire team had this. Today you'll learn how to share the power.
Three ways to share
There are three models for giving your team access to AI agents:
Shared instance — one agent, multiple users. Everyone talks to the same agent. It knows the team's context, shared documents, and group preferences. Best for small teams with shared workflows.
Cloned configuration — everyone gets their own agent, but they start from the same base setup. Same tools, same workflows, same custom instructions. Each person's agent then learns their individual preferences. Best for larger teams.
Team + personal — a shared team agent handles group tasks (standup digests, team reports) while each person also has a personal agent for individual work. Best of both worlds.
Knowledge Check
Which sharing model works best for a 20-person team where everyone has different workflows?
A
No sharing — everyone builds from scratch
B
Cloned configuration — same base setup, individual personalization
C
Team + personal — but only for managers
D
Shared instance — one agent for everyone
Cloned configuration gives everyone the same starting point (saving setup time) while letting each agent learn individual preferences. A shared instance would be overwhelmed with 20 different people's preferences and requests.
Setting up a team agent
Let's see what a shared team agent looks like:
Post the daily standup digest.
📊 **Standup Digest — Thursday, Feb 19**
**Sarah**: Completed auth refactor, starting API docs
**Marcus**: Reviewing PRs, will finish by EOD
**Priya**: Client design review at 2pm, assets to Marcus by EOD
**Jake**: Dashboard fix deployed, picking up performance ticket
**Liam**: Out sick today — no update
**Nina**: Sprint planning prep for Monday
🚩 **Blockers**: Marcus waiting on Priya's assets (ETA: today EOD)
📅 **Upcoming**: Sprint planning Monday 10am
Posted to #engineering-leads ✓
↻ Replay conversation
Permissions matter
When multiple people share an agent, permissions become critical:
Who can see what? — The agent should respect data boundaries. Sarah shouldn't see Marcus's personal emails through the team agent.
Who can do what? — Can everyone trigger workflows, or only admins? Can anyone change custom instructions, or is that locked?
What's shared vs. personal? — Team memory (project details, client info) is shared. Personal memory (individual preferences, private notes) stays private.
Knowledge Check
Why is permission scoping important when sharing an agent?
A
To make the agent faster
C
To prevent people from accidentally accessing each other's private data or modifying shared configurations
D
Permissions aren't important for small teams
Without permission scoping, any team member could read anyone else's email through the agent, or change a workflow that the whole team depends on. Scoping ensures everyone has access to what they need — and nothing they shouldn't.
Cost splitting
When your team shares an agent (or runs from the same infrastructure), costs can be distributed:
Shared instance: one bill, split across the team. Simple.
Cloned configurations: each person pays for their own API usage, but shares server costs.
Company-provided: many companies now provide AI agent budgets as part of their tooling. Your agent is a work tool, just like Slack or Jira.
With the cost optimizations from Day 19, a team of 10 can run agents for less than the price of one SaaS tool subscription.
What would it cost to give agents to my 10-person team?
Here's the breakdown with optimized settings:
🖥️ **Hosting**: $10/month (shared VPS)
🤖 **API costs**: ~$70/person/month with model tiering
💰 **Total**: ~$710/month for 10 agents
That's $71/person/month — less than most SaaS tools. And each person saves 2-3 hours daily.
Compare that to hiring an assistant at $4,000/month who can only help one person at a time.
↻ Replay conversation
Knowledge Check
At $71/person/month, what's the ROI if each team member saves 2 hours per day?
A
Impossible to calculate
B
Massive — 40+ hours saved per month per person, far exceeding the cost
D
Negative — it's too expensive
At 2 hours saved per day, that's 40+ hours per month. Even at a modest hourly value, the time saved far exceeds $71/month. And unlike a human assistant, the agent works 24/7 and never takes vacation.
Getting your team started
The fastest way to onboard your team:
1. Export your agent's configuration (tools, custom instructions, workflows)
2. Set up a shared server or distribute the config to individual installs
3. Have each person connect their own email, calendar, and Slack
4. Share this course — they'll be up to speed in 28 days
The configuration is portable. What took you weeks to build, your teammates can start using in an afternoon.
Final Check
What's the fastest way to get a teammate started with OpenClaw?
A
Give them your login credentials
B
Export your configuration and let them import it — then personalize their own preferences
C
Have them build everything from scratch
D
Buy them a subscription to a different AI service
Your configuration is exportable. Share it, and your teammate gets all your tools, workflows, and custom instructions instantly. They just need to connect their own accounts and let the agent learn their individual preferences.
👥
Day 27 Complete
"One agent changed how you work. Ten agents will change how your team works."
Tomorrow — Day 28
The OpenClaw Community
You're not alone — there's a massive community building, sharing, and improving OpenClaw every day.