Skip to main content
Komo’s Cloud Compute Environment provides each agent with an isolated virtual computer in the cloud—complete with persistent storage, terminal access, and computational resources. This enables agents to run complex, long-duration tasks in the background without relying on your local machine.

​
What is the Cloud Compute Environment?

The Cloud Compute Environment is a fully-featured, sandboxed Linux environment that runs in Komo’s secure cloud infrastructure. Each agent operates in its own isolated compute instance with:
  • Persistent file system - Files and data persist across agent executions
  • Terminal access - Execute shell commands, run scripts, install software
  • Computational resources - CPU, memory, and storage dedicated to agent workloads
  • Network connectivity - Access to internet, APIs, and external services
  • 24/7 availability - No downtime, always ready for scheduled or triggered execution
Unlike local execution which requires your computer to remain online, cloud compute runs independently—enabling true autonomous, always-on AI operations.

​
Why Cloud Compute Matters

​
The Problem with Local Execution

Traditional automation tools require your local computer to:
  • Remain powered on and connected
  • Have sufficient resources available
  • Handle interruptions (sleep mode, network loss, crashes)
  • Manage security of local credentials and data
This creates operational limitations:
  • Can’t run overnight or multi-day tasks reliably
  • Computer resources compete with your daily work
  • Security risks from storing credentials locally
  • No scalability—limited by single machine capacity

​
The Cloud Compute Solution

Komo’s Cloud Compute Environment eliminates these limitations: Always-On Operations
  • Run tasks 24/7 without your computer online
  • Schedule agents for overnight or weekend execution
  • Multi-day workflows complete reliably
Resource Isolation
  • Dedicated compute resources per agent
  • No impact on your local machine performance
  • Scale to hundreds of parallel agents
Enhanced Security
  • Isolated sandbox environment per user
  • No local access to your computer or files
  • Credentials managed in encrypted cloud storage
  • Complete audit trail of all operations
Scalability & Reliability
  • Enterprise-grade infrastructure
  • 99.9% uptime guarantee
  • Automatic failover and recovery
  • Horizontal scaling for parallel execution

​
How It Works

​
1. Automatic Provisioning

When you create an agent, Komo automatically provisions:
  • Isolated Linux container
  • Persistent file system (/workspace directory)
  • Network connectivity
  • Computational resources
  • Pre-installed tools and libraries
No configuration required—agents start working immediately.

​
2. Persistent Workspace

Each agent has a dedicated /workspace directory that persists across executions: 
/workspace/
├── data/           # Store datasets, files, extracted data
├── scripts/        # Custom scripts and automation
├── outputs/        # Generated reports, documents, artifacts
├── cache/          # Cached data for faster subsequent runs
└── temp/           # Temporary files from current session
Temporary File Management: All temporary files created during agent execution (screenshots, downloads, intermediate data) are automatically saved to /workspace/temp/ with limited retention:
  • Standard retention: 1 day from creation
  • Automatic cleanup: Files deleted after retention period expires
  • Enterprise custom retention: Contact [email protected] for custom retention policies (30/60/90 days or indefinite storage)

​
3. Tool Availability

Pre-installed tools in every compute environment: Programming Languages:
  • Python (with pip)
  • Node.js (with npm)
  • Ruby
  • Go
Data Processing:
  • Pandas, NumPy (Python)
  • Excel/CSV processors
  • JSON/XML parsers
  • Database clients (PostgreSQL, MySQL, MongoDB)
Web Automation:
  • Browser automation libraries
  • HTTP clients
  • Web scraping tools
  • API integration utilities
System Tools:
  • Git (version control)
  • wget/curl (downloads)
  • Text processors (sed, awk, grep)
  • Compression utilities
Custom Software: Clean Shot 2026 01 06 At 14 59 06@2x You could manage and config those tool availaibility in the agent level at the Tools section of the playbook’s sidebar:
  • Browser automation
  • File Manager
  • Terminal tools
  • Deploy tools
  • Image processing tools
  • Data provider tools (Linkedin, Pitchbook etc)
  • Image Editor
  • (Enterprise) Other custom tools to pre-install

​
4. Execution Model

Background Execution:
  • Agents run independently of your connection
  • You can close your browser—agents continue running
  • Check status anytime via Activity Monitor
Long-Running Tasks:
  • Tasks can run for hours or days
  • No timeout limitations
  • Progress tracked in real-time
Parallel Execution:
  • Run multiple agents simultaneously
  • Each in isolated environment
  • No resource contention

​
Real-World Examples

​
Example 1: Daily Market Research (Always-On)

Scenario: Investment team needs daily competitive intelligence Agent Configuration:
  • Schedule: Every weekday at 6 AM EST
  • Environment: Cloud compute (runs even when team offline)
Workflow: Daily at 6 AM (cloud agent runs autonomously):
  1. Research 20 competitor companies
  2. Download latest financial filings
  3. Extract key metrics to /workspace/competitors.db
  4. Analyze pricing changes vs. yesterday’s data
  5. Generate trend report
  6. Post summary to Slack #competitive-intel
  7. Team sees report when they arrive at 9 AM
Result:
  • Runs reliably every morning
  • No team member needs to be online
  • 90 days of historical data accumulated in /workspace
  • Complete competitive intelligence ready each morning

​
Example 2: Multi-Day Data Processing

Scenario: Data team needs to process 10TB dataset Why Cloud Compute:
  • Process would take 36 hours
  • Can’t keep laptop running for 36 hours straight
  • Need reliable, uninterrupted execution
Workflow: Agent starts processing:
  • Hour 0: Download dataset chunks to /workspace/data/
  • Hour 6: Process first 25% of data
  • Hour 12: Process 50% complete (your computer can sleep/shutdown)
  • Hour 24: Process 75% complete (agent still running in cloud)
  • Hour 36: Processing complete, results in /workspace/outputs/
  • Agent notifies: “Processing complete. Results ready for download.”
Result:
  • Completed reliably over weekend
  • No local computer resources used
  • Results ready Monday morning

​
Example 3: Parallel Research Execution

Scenario: Research team needs analysis of 500 companies Why Cloud Compute:
  • Sequential processing would take days
  • Need to deploy multiple agents simultaneously
  • Require reliable, scalable execution
Workflow: Deploy 50 separate agents in cloud compute:
  • Each agent: isolated environment with own /workspace
  • Each agent: researches 10 companies independently
  • All agents: execute simultaneously
  • Results: each agent produces report in their /workspace
  • Aggregation agent: collects all reports into master dataset
  • Time: 2 hours (vs. 100+ hours sequential)
Result:
  • 50x faster through multi-agent parallelization
  • Each agent runs independently
  • No local resources used
  • Scalable to 1000s of companies

​
Example 4: Scheduled Backup & Monitoring

Scenario: DevOps team needs continuous infrastructure monitoring Agent Configuration:
  • Schedule: Every 15 minutes, 24/7
  • Environment: Cloud compute (never stops)
Workflow: Every 15 minutes (cloud agent checks):
  1. Query production database health
  2. Check API response times
  3. Monitor error rates
  4. Compare against baselines in /workspace/metrics/
  5. If anomaly detected: alert #devops-alerts Slack
  6. Update /workspace/metrics/historical.db
  7. Generate weekly trend report (every Monday)
Result:
  • 24/7 monitoring without dedicated server
  • Historical data accumulated automatically
  • Anomalies detected in real-time
  • Zero maintenance required

​
Security & Isolation

​
Sandboxed Execution

Each agent runs in a completely isolated container: Network Isolation:
  • No access to other agents’ environments
  • No access to Komo’s internal systems
  • Outbound internet access only (controlled)
File System Isolation:
  • Each agent has separate /workspace
  • No cross-agent file access
  • No access to host system
Process Isolation:
  • Agents cannot see other agents’ processes
  • Resource limits enforced
  • Prevents one agent from affecting others

​
No Local Access

Critical Security Benefit:
  • Agents NEVER access your local computer
  • No local files, no local browsing history
  • No local credentials or cookies
  • Complete separation from your personal system
This means:
  • Your local data remains private
  • Local security posture unchanged
  • No malware risk to local machine
  • No performance impact on local computer

​
Credential Management

Secure Storage:
  • API keys and credentials stored encrypted
  • Separate encrypted vault per user
  • Never logged or exposed in plain text
Access Control:
  • Credentials accessible only to your agents
  • Time-limited access tokens
  • Revocable at any time
Audit Trail:
  • Complete log of credential usage
  • Which agent accessed what, when
  • Export logs for compliance

​
Managing Cloud Compute

​
File Management

Download Files:
  • Download any file from /workspace to local machine
  • Bulk download entire directories
  • Access temporary files within retention period
Upload Files:
  • Upload datasets, scripts, configuration files
  • Agents can access uploaded files immediately
Storage Management:
  • View storage usage per agent
  • Clean up old files manually
  • Archive historical data
  • Temporary files auto-deleted per retention policy

​
Best Practices

​
Use Cloud Compute For:

✅ Scheduled & Triggered Tasks
  • Agents running on schedule (daily/weekly/monthly)
  • Event-driven workflows (when X happens, do Y)
  • Background processing without human supervision
✅ Long-Running Operations
  • Tasks taking hours or days
  • Large-scale data processing
  • Multi-step workflows with dependencies
✅ Parallel Execution
  • Processing hundreds/thousands of items
  • Multi-agent coordination
  • High-volume operations
✅ Always-On Monitoring
  • 24/7 surveillance and alerting
  • Continuous data collection
  • Real-time anomaly detection

​
Common Questions

Q: Does cloud compute access my local computer? A: No. Cloud compute runs entirely in isolated cloud infrastructure. Zero access to your local machine, files, or network. Q: What happens if my internet disconnects? A: Agents continue running in the cloud. Reconnect anytime to check status and results. Q: Can agents run while I’m offline? A: Yes. That’s the primary benefit. Schedule agents to run overnight, weekends, anytime—no need to be online. Q: How is cloud compute different from cloud browser? A: Cloud compute is the full Linux environment (terminal, file system, tools). Cloud browser is the web browser running within that environment. Q: Is my data secure in cloud compute? A: Yes. Isolated containers, encrypted storage, no cross-user access, full audit trails. Enterprise-grade security. Q: What are the resource limits? A: Varies by plan. Standard: 4 CPU cores, 16GB RAM, 100GB storage per agent. Enterprise: customizable. Q: Can I install custom software? A: Yes. This is an Enterprise feature, please contact sales team for requests. Q: Can multiple agents share data? A: Within same workspace, yes. Agents can read/write shared /workspace directory. Q: How long are temporary files kept? A: Standard: 1 day from creation, then automatically deleted. Enterprise: custom retention available—contact [email protected]. Q: Can I access temp files from previous sessions? A: Yes, if within the 1-day retention window. After that, they’re automatically cleaned up to save storage.