Skip to main content

Confirmation Modes

Bridge has two modes that control how write actions are handled. Toggle between them using the Safe / Auto switch in the Bridge header.


Safe Mode (Default)

In Safe Mode, every write action shows a preview card before executing. You see exactly what Bridge plans to do and must click Confirm to proceed.

You: "Create an agent called researcher with web_search enabled using gpt-4o"

Bridge: I'll create this agent for you:

┌─────────────────────────────────────────────────────────────┐
│ ⚙️ Create Agent │
│ │
│ Name: researcher │
│ Model: gpt-4o │
│ Capabilities: web_search ✅ │
│ Temperature: 0.7 │
│ Max Tokens: 1024 │
│ System Prompt: "You are a research assistant..." │
│ │
│ API Key: [Select below] │
│ ○ My OpenAI Key │
│ ● My Claude Key (not compatible with gpt-4o) │
│ │
│ [✓ Confirm] [✏️ Modify] [✗ Cancel] │
└─────────────────────────────────────────────────────────────┘

When to use Safe Mode:

  • When learning Bridge — see exactly what each command does
  • When making important configuration changes
  • When managing production workflows
  • When you're not sure of the exact effect of a command

Auto Mode

In Auto Mode, non-destructive write actions execute immediately without a preview card. Only destructive actions (delete, reject) still require confirmation.

You: "Create an agent called researcher with web_search enabled using gpt-4o"

Bridge: ✅ Done! I've created the "researcher" agent with:
- Model: gpt-4o
- web_search capability enabled
- Default system prompt for research tasks

You'll need to select the API key in the Agent Editor.
Want me to do anything else with this agent?

When to use Auto Mode:

  • When you're an experienced Bridge user who understands what commands do
  • When rapidly iterating on agent configs (create, test, adjust, repeat)
  • When doing bulk configuration work
Auto Mode still confirms deletes

Even in Auto Mode, deleting agents, workflows, or other resources always shows a confirmation prompt. This is a safety feature that cannot be disabled.


Preview Card Anatomy

When a preview card appears (Safe Mode, or destructive action in Auto Mode):

┌─────────────────────────────────────────────────────────────┐
│ [Action Icon] [Action Title] [Action] │
│ │
│ [Field label]: [Value] │
│ [Field label]: [Value] │
│ [JSON Preview or structured config] │
│ │
│ [✓ Confirm] [✏️ Modify] [✗ Cancel] │
└─────────────────────────────────────────────────────────────┘
ElementDescription
Action Icon⚙️ Create, ✏️ Update, 🗑️ Delete, ▶ Execute, 🔗 Connect
Action TitlePlain English description of what's happening
FieldsStructured view of what will be created/changed
ConfirmProceeds with the action as shown
ModifyOpens an inline editor to adjust values before confirming
CancelCancels the action entirely

Modifying Before Confirming

Click Modify to adjust the proposed action:

┌─────────────────────────────────────────────────────────────┐
│ ✏️ Modify Agent Creation │
│ │
│ Name: [researcher_____] │
│ Model: [gpt-4o_________] ← Change here │
│ Temp: [0.7__] │
│ web_search: [✓] │
│ web_scraper: [☐] │
│ │
│ System Prompt: │
│ [You are a research assistant...____ │
│ ___________________________________] │
│ │
│ [✓ Create with these settings] [✗ Cancel] │
└─────────────────────────────────────────────────────────────┘

After modifying, Bridge creates the agent with your adjusted values.


Destructive Action Confirmation

Destructive actions (delete, cancel, reject) always require explicit confirmation, regardless of mode:

You: "Delete the old-research-agent"

Bridge: ⚠️ Delete Agent

Are you sure you want to permanently delete:

Agent: old-research-agent
Model: gpt-4o
Used in: [research-pipeline workflow]

⚠️ This workflow will have a broken node after deletion.

[🗑️ Yes, delete permanently] [✗ Cancel]

Bridge also warns you if the deletion will break dependent resources (like workflows using that agent).


Switching Modes

Click the Safe / Auto toggle in the Bridge header. The mode is:

  • Saved per user account (persists across sessions)
  • Per-project (you can use Safe in one project and Auto in another)
  • Applied immediately — the current conversation switches modes

Mode Recommendation

User TypeRecommended Mode
New to BridgeSafe — see what each command does
Learning agent/workflow creationSafe — review each configuration before creating
Power user, production projectSafe — avoid accidental changes
Power user, development projectAuto — move faster
Bulk configuration sessionsAuto — create many resources quickly