Back
— · 4 Min read

Service Agents Management: Your Admin Control Panel for AI Agents

Managing AI agents in production used to mean SSH, terminal commands, and hoping you don't typo a process ID. Today we're introducing Service Agents Management - a full admin control panel for your workspace agents, right in OpenAgents Studio.

The Problem

Running agent networks means managing multiple agent processes:

  • Starting agents: openagents agent start config.yaml for each one
  • Checking status: ps aux | grep python or custom scripts
  • Viewing logs: tail -f workspace/logs/agents/my_agent.log
  • Editing configs: Open files in your editor, save, restart manually

That's a lot of context switching and terminal juggling, especially when you have multiple agents.

The Solution: Point-and-Click Agent Management

Service Agents Management puts everything in one place:

See Everything at a Glance

The sidebar shows all your agents with live status indicators:

  • Green dot: Running smoothly
  • Gray dot: Stopped
  • Red dot: Error (with message)
  • Yellow pulse: Starting or stopping

Quick counts tell you how many agents are in each state.

Control with a Click

No more typing commands. Select an agent and:

  • Start: Launch a stopped agent
  • Stop: Gracefully shut down a running agent
  • Restart: Quick stop-and-start

The UI shows loading states and success/error notifications so you know what's happening.

Watch Logs in Real-Time

The Logs tab gives you a full-height log viewer with:

  • Live updates: Polls every 2 seconds for running agents
  • Filtering: Show only INFO, WARN, or ERROR
  • Auto-scroll: Follows new logs automatically
  • Clear: Reset the view when it gets cluttered

No more tail -f in a separate terminal.

Edit Code in the Browser

Here's the game-changer: edit your agent source code directly in Studio.

We've integrated Monaco Editor (the same editor that powers VS Code) with:

  • Full syntax highlighting for Python and YAML
  • Line numbers and code folding
  • Bracket matching
  • Dark mode support

Make changes, hit Save, and you'll get a backup (.bak file) before we overwrite. If the agent is running, we'll prompt you to restart.

Admin-Only Access

This is powerful functionality, so it's locked down:

  • Only administrators can access Service Agents Management
  • Non-admins see an "Access Denied" message
  • Agent groups control who's an admin

Your agents are safe from accidental (or malicious) changes by unauthorized users.

How It Works

Agent Discovery

Drop .yaml or .py files in workspace/agents/ and they're automatically discovered:

workspace/
└── agents/
    ├── support_bot.yaml      # YAML agent
    ├── data_processor.py     # Python agent
    └── research_agent.yaml   # Another YAML agent

For YAML files, we read the agent_id field. For Python files, we look for default_agent_id or agent_id variables.

Process Management

The AgentManager handles the heavy lifting:

  • Start: Spawns the process, captures stdout/stderr
  • Stop: Sends SIGTERM, waits 5 seconds, then SIGKILL if needed
  • Monitor: Watches for crashes and updates status

Logs go to workspace/logs/agents/{agent_id}.log with session markers.

API-Driven

Everything works through HTTP APIs, so you can:

# List all agents
curl http://localhost:8700/api/agents/service
 
# Start an agent
curl -X POST http://localhost:8700/api/agents/service/my_bot/start
 
# Get source code
curl http://localhost:8700/api/agents/service/my_bot/source
 
# Update source code
curl -X PUT http://localhost:8700/api/agents/service/my_bot/source \
  -H "Content-Type: application/json" \
  -d '{"content": "agent_id: my_bot\n..."}'

Build your own tooling on top if you want.

Real-World Workflow

Here's how I use it:

  1. Morning check: Open Studio, glance at the sidebar - all green? Good.

  2. Something's wrong: Red dot on support_bot. Click it, check the Logs tab. Oh, the API key expired.

  3. Quick fix: Source Code tab, update the key, Save.

  4. Restart: "Restart Now" when prompted. Watch the status go yellow, then green.

  5. Verify: Back to Logs tab, see the agent starting up successfully.

All without leaving the browser.

Try It Now

Update to the latest OpenAgents:

pip install -U openagents

Make sure you have agents in your workspace:

# workspace/agents/hello_bot.yaml
agent_id: hello_bot
model_name: gpt-4o
system_prompt: You are a friendly bot that says hello.

Log in as an admin and navigate to Service Agents in Studio. You'll see your agents ready to manage.

What's Next

This is v1 of Service Agents Management. On the roadmap:

  • Agent creation: Create new agents from templates
  • Batch operations: Start/stop multiple agents at once
  • Metrics: CPU/memory usage per agent
  • Alerts: Notifications when agents crash

Feedback Welcome

How do you manage your agents today? What features would make your life easier?

Let us know:

Happy managing!

The OpenAgents Team