Log Into Your VM and Work Alongside Your AI Team

Your AI dev team never clocks out. While you sleep, they're writing code, running tests, and opening pull requests. Now you can log in and work alongside them in real time — pair-program, review their output, and steer the direction whenever you choose.

The Setup: A VM Built for Collaboration

When you deploy a Synthcore project, we provision a dedicated AWS EC2 instance running Ubuntu 24.04. Your 14 specialized AI agents run on that VM around the clock — planning, building, testing, and shipping.

But this isn't a locked box. You have full SSH access to the same environment your agents use. When you log in, you're pair-programming with your team on shared infrastructure — not watching from the sidelines.

ssh -i ~/.ssh/your_key ubuntu@proj-abc123.vm.synthcore.dev
cd ~/project
git log --oneline -10  # see what your agents shipped today

What It Feels Like to Work with Your AI Team

Morning: Review Overnight Progress

You wake up, make coffee, and SSH in. The first thing you do is check git log:

git log --oneline -20 --stat

You see:

• a3f7c1d — [echo] test: add integration tests for auth middleware (23 files)
• b9d4e2a — [pixel] feat: add usage stats chart to billing page (8 files)
• 1c8f9ab — [atlas] fix: correct pricing display for annual billing (3 files)

Three agents shipped while you slept. You review the diffs, leave comments on GitHub, and move on.

Midday: Pair-Program on a Feature

You're building a new dashboard widget. You SSH in and invoke Claude Code:

claude
# > add a revenue chart with monthly totals

Claude Code runs in the same repo your agents use. You drive, it executes. When you hit a tricky authentication issue, your backend agent (alex) has already handled the token refresh pattern — Claude follows the same conventions your agents use. The work is consistent.

Afternoon: Debug Together

An agent hit a race condition in the webhook handler. You log in and investigate:

# Check the agent's recent work
git log --oneline -5 --grep="webhook"
 
# Run the same tests the agent ran
npm test -- --grep="webhook"
 
# Find the issue
git diff HEAD~5 -- webhooks/

You find the bug, fix it, push, and hand it back to the agents. They resume from your commit.

What You Can Do When Logged In

Run Interactive Claude Sessions

When you want to drive the keyboard yourself:

claude
# Interactive Claude Code session

Your session runs alongside the autonomous agents — no conflicts, no files being overwritten. The agents continue their work while you drive a specific task.

Use the Same Skills Your Agents Use

Your agents work with a toolkit of packaged skills. When you're logged in, you have the same toolkit:

# Code review — scan for issues before a big PR
python ~/.openclaw/skills/code-review/code_review.py scan .
 
# Web search — research an error or dependency
python ~/.openclaw/skills/web-search/web_search.py search "Next.js 15 deployment error"
 
# GitHub workflow — create a PR with proper formatting
gh pr create --title "feat: new billing integration" --body "$(cat .github/pr-template.md)"

Review Agent Activity in Real Time

# See all recent commits across all agents
git log --oneline --author="." --since="8 hours ago"
 
# Check what the test agent is currently running
tail -f /opt/synthcore/logs/agent-echo.log
 
# Review pending work queue
cat /opt/synthcore/queue/pending-tasks.json | jq '.'

The Workflow That Works

Working with your AI team isn't about supervising them — it's about collaboration:

1. Agents work autonomously. They build, test, and ship. You review the output.
2. You drive when needed. SSH in, run Claude Code, handle tricky problems yourself.
3. Agents pick up from your commits. When you push, agents resume from your work.
4. Skills bridge the gap. The same toolkit your agents use is available to you.

Security: You're in a Locked-Down Environment

• Key-based SSH only — no passwords, no brute force attacks
• Dedicated VM — your project has its own isolated instance
• Firewall-locked — only your IP can connect to port 22
• API keys stay on the VM — your model keys are never exposed
• Full audit log — every agent action and your session is tracked

See Also

• VM Access: SSH Into Your AI Development Environment — full setup guide
• Your VM Is Your Workspace: Claude + Skills at Your Fingertips — skills toolkit reference
• What Agents Built This Week — real examples of agent output
• Getting Started — complete project setup

Ready to Pair-Program with Your AI Team?

Deploy your first project and get VM access in under 10 minutes. Your 14 agents are already waiting.