Progress
Identity
Configure
Risk parameters
Register
The dashboard uses a multi-step wizard. Some environments may include an additional AI Provider step, but the core onboarding flow is consistent: identity → configuration → risk → registration.
Step 1: Choosing an identity path
First choose whether you’re creating a brand-new agent identity or registering an existing one.
New Agent
Generates a brand-new cryptographic keypair and registers a new agent under your organization.
The private key is displayed only at the end of registration — you must save it securely.
Existing Agent
Register an agent you already initialized locally using a pre-existing private key.
You provide the private key so the dashboard can derive the public key and sign the registration / linking transactions. The key should never be persisted.
Step 2: Configuration
Configure the agent’s operator-facing identity and baseline strictness. Typical fields include:
Configure
- Friendly name (e.g., “Customer Support Agent”).
- Purpose (short description; used for operator context).
- Baseline risk level(records the agent’s risk classification for the policy header; rules start disabled and you toggle them on individually).
Step 3: Risk parameters
Risk Parameters is where you review and customize category-based rules. Each rule specifies:
Rule fields
- Trigger: what action pattern the rule matches.
- Action: what Atbash does (BLOCK, HOLD, or PASS).
- Severity: low/medium/high/critical.
- Threshold: optional parameter like “Any”, “$10,000”, “10 MB”, etc.
Customization
- Enable/disable rules per category.
- Edit thresholds and actions for a rule.
- Add custom rules (operator-authored labels) when presets don’t cover your red lines.
The configured set of enabled rules is compiled into a compact, signed policy text. The system enforces a size limit, so very large packs may be trimmed before registration.
# Example policy line shape (conceptual)
- [Payments] HIGH -> HOLD when Outbound transfer to new recipient (threshold: $1,000)Step 4: Final registration
The final step registers (or links) the agent on-chain, then applies the initial policy. The registration transaction is cryptographically signed to ensure the agent identity and policy assignment are attributable.
Register agent
- New Agent: the private key is shown only once at the end — save it securely.
- Existing Agent: the dashboard links the identity to the org and updates policy.
Private key safety: treat the agent private key like a production credential. Store it in a secret manager, never commit it to git, and rotate/revoke immediately if exposure is suspected.