COMING SOON — Join the waitlist for early access
Bouncer

Trust Levels

Trust levels determine how much independence each AI agent has. You set them per agent, and you can change them at any time.

The Four Levels

Locked

The agent can only prepare drafts. Every single action requires your explicit approval.

Best for:

New agents you haven't worked with yet
Sensitive areas (financial transactions, legal communications)
When you're first setting up and want to see how agents behave

What happens:

Agent drafts content and places it in the Queue
Nothing is sent, executed, or modified until you approve
You see every detail before anything happens

Supervised

The agent can handle routine, low-risk tasks independently. Higher-risk actions still need approval.

Best for:

Agents you've worked with for a few weeks and trust with basic tasks
Most users' default setting

What's automatic:

Reading and analyzing data
Internal note-taking and organization
Low-risk drafting

What needs approval:

Outbound messages (email, SMS)
Modifications to contact or property data
Financial categorizations
Scheduling actions that affect others

Trusted

The agent handles most tasks independently. Only unusual or high-impact actions need approval.

Best for:

Agents that have proven reliable over time
Areas where speed matters (lead follow-up, time-sensitive responses)

What's automatic:

Everything in Supervised, plus:
Routine outbound messages following approved templates
Standard data updates
Scheduled actions within normal parameters

What needs approval:

First-time recipients (contacts you haven't messaged before)
Actions above a cost threshold you set
Unusual patterns (much higher volume than normal)

Full

The agent operates autonomously within its domain. Actions are logged but not held for approval.

Best for:

Mature workflows you've fully validated
High-volume operations where manual approval creates bottlenecks

What's automatic:

All actions within the agent's domain
Everything is logged in the Activity feed

What still happens:

Cost limits are still enforced — the agent pauses when credits run low
The kill switch still works — you can stop everything instantly
Activity is fully auditable — nothing happens without a record

Changing Trust Levels

1.Go to Bouncer > Trust Levels
2.Select the agent you want to adjust
3.Choose the new trust level
4.Confirm the change

Changes take effect immediately. If you lower a trust level, any pending actions from that agent are re-queued for approval.

Start all agents at Locked or Supervised. After reviewing 20-30 approved actions from an agent, consider moving to Trusted if you're satisfied with the quality. Move to Full only for workflows you're completely confident in.

Example: Trust Level Progression

Here's a realistic timeline for a new user:

Week 1 (Locked): You review every email, every data change, every draft. You notice the Leasing Agent writes great tenant responses but the Campaigns Agent needs more coaching on your tone.

Week 3 (Supervised): You move the Leasing Agent to Supervised. It now handles routine inquiries and data lookups on its own. You still approve all outgoing messages. The Campaigns Agent stays at Locked while you keep editing its drafts.

Month 2 (Trusted): The Leasing Agent moves to Trusted. It sends routine responses (showing confirmations, maintenance updates) using your approved templates. You only review first-time contacts and high-dollar items. The Campaigns Agent moves to Supervised.

Month 4+ (Full for select agents): Your Leasing Agent runs at Full for a property with 20 units — manual approval at that volume would be a bottleneck. Your Bookkeeping Agent stays at Supervised because you prefer to review financial categorizations yourself.