Agents you can audit
Agents work inside limits you set. Every draft is checked before it sends and logged after, and your data is never used to train models for other customers.
People stay in control
Agents earn autonomy the way a new hire does. Every agent moves through the same three levels, and you decide when it moves up.
Level 1: Shadow
The agent drafts but never sends. Every new agent starts here, so you can read what it would have said.
Level 2: Approval
The agent drafts and a person signs off. Approve, edit, snooze, discard or reassign each draft.
Level 3: Autonomous
The agent sends on its own unless a rule holds the draft. You choose which agents and accounts get here.
One approval queue
Drafts that are low-confidence, escalated or held by a guardrail wait for a person, with their sources and reasoning.
Guardrails outrank rules
Auto-approve rules release the routine drafts. A guardrail can never be overridden by a rule.
Hard limits at every level
Limits on pricing below the approved band, delivery promises, contract changes, refunds and roadmap promises apply at every autonomy level.
Take over anytime
Step into any thread mid-draft, or stop an auto-send countdown. A held draft explains why it was held.
Every action is accountable
You can see who did what, and why: the person or the agent, the sources behind it and the rule that let it through.
An audit log of every action
Every draft, edit, approval and send is written down with the sources the agent used.
Cited sources
Each draft carries the knowledge it relied on. Agents can be required to cite a source for every claim.
Confidence and reasoning
“Why this draft?” shows the steps the agent took. Agent runs list the rules it checked along the way.
Undo where it counts
Approve, snooze, discard and reassign can all be undone. Only a sent message can’t.
Your data stays yours
Mokanata holds your customer conversations, so it treats them that way: protected while it has them, and yours to take.
Encrypted in transit and at rest
Conversations and records are encrypted on the way in and while they are stored. So are the secrets of connected apps.
Never used to train models for other customers
Your conversations and records stay yours. Pay as you go can also bring custom models.
Export anytime
Contacts, deals and conversation history can be exported at any time.
Verified webhooks
Incoming webhooks are checked by HMAC signature or a secret header, and their secrets can be rotated.
Access and identity
Decide who can read, reply and change what, down to the channel account.
Roles and permissions
Owner, Admin, Member and Viewer, or custom roles built in a permission editor that covers every area.
Channel access per role
Give a role all accounts or a chosen set, at View, Reply or Manage, and see which accounts nobody can reply from.
Two-step verification
An authenticator app with recovery codes. Changing your login email is confirmed by email.
Session control
See active sessions and sign out everywhere else.
Scoped API keys
Read-only or read-write, with an expiry. The secret is shown once, and read-only keys see only the MCP server’s read tools.
On Pay as you go
Pay as you goFor teams that sign in through an identity provider, or whose data has to stay in one region.
- SSO / SAML
- SCIM
- Audit log export
- Custom models
- Data residency
- 99.9% SLA
Security questions?
Ask about how Mokanata handles your data, access or agents. Write to sales@mokanata.com and someone on the team will answer.
Start in shadow mode today
Connect a channel, let an agent draft for a week and read what it would have sent. Nothing goes out until you say so.