Dashboard Guide

Page-by-page walkthrough of the Aira dashboard — what each section does and how to use it.

Everything you can do in the Aira dashboard, organized by sidebar section. Each page links back to the relevant API reference if you want to do the same thing programmatically.

Platform

Overview

The landing page after login. Shows four metric cards (actions notarized, agents registered, evidence packages, usage), plus quick-access tables for recent actions, agents, and cases.

If your account is new, each card prompts you to take the first step: notarize an action, register an agent, or submit a case.

Natural-language interface over your organization's data. Type a question like "what did lending-agent do last week?" or "is any agent drifting?" and the assistant queries your actions, agents, policies, cases, receipts, usage, drift alerts, audit logs, compliance snapshots, and settlements.

Pick any supported model from the dropdown (Anthropic, OpenAI, Google). The model must be configured by an admin under Models > Preferences before chat works.

Chat conversations are ephemeral — they reset when you leave the page. If you need a permanent record, copy the answer or export the referenced data.

Agents

Agent List

Table of all registered agent identities. Each row shows the agent slug, display name, status (active/inactive), and registration date.

Register a new agent using the button in the top-right. You'll assign a slug (permanent, URL-safe identifier), display name, and optional capabilities list. Once registered, the agent gets a W3C DID and can be referenced in authorize() calls.

Agent Detail

Click any agent to see its full profile:

Identity — slug, display name, description, capabilities, DID document
Versions — published version history with config snapshots
Will — succession plan for when the agent is decommissioned
Reputation — verifiable credential + computed reputation score
Drift — behavioral baseline and KL divergence alerts (separate tab)

Key actions: edit metadata, publish a new version, rotate keys, decommission (issues a cryptographic death certificate), configure will.

Agent Drift

Per-agent behavioral drift analysis. Shows the agent's baseline distribution and any divergence alerts. If the agent starts doing something new (new action types, new patterns), drift detection fires an alert.

Baselines are created automatically on the agent's first actions, or manually via POST /api/v1/agents/{id}/drift/baseline.

Actions

Table of every action your agents have taken. Each row shows action UUID, type, agent, status, legal hold, and date.

Status lifecycle:

authorized — policy engine approved, agent can proceed
pending_approval — held for human review
approved — human approved the held action
denied_by_policy — policy blocked the action (still billed)
notarized — agent reported the outcome, receipt minted
failed — agent reported a failure

Key actions: filter by type/agent/status, set or release legal hold on any action, click through to the action detail page where you can see the full receipt, policy evaluation, and signed payload.

Legal hold prevents an action's data from being deleted or modified during an investigation. It's a flag, not encryption — the data is still readable, just protected from cleanup processes.

Decisions

Cases

Multi-model consensus runs. Each case sends the same question to multiple AI models and computes a consensus decision with a disagreement score.

Create a case using the top-right button. Choose 2-5 models, provide the decision context, and submit. The case runs all models in parallel and returns:

Individual model decisions + reasoning
Consensus decision (approve/deny/review)
Confidence and disagreement scores
Whether human review was triggered

Cases with high disagreement (configurable threshold) automatically enter pending_review and appear in the approvals queue.

Receipts

Cryptographic receipts tied to cases and actions. Each receipt row shows the decision, confidence, human review flag, and creation date.

Click any receipt to see the full cryptographic proof: payload hash, Ed25519 signature, signing key ID, and the complete signed payload JSON. You can verify the signature, export as JSON, or download as PDF.

Receipts are generated automatically — you never create them manually. Every successful notarize call and every completed case produces a receipt.

Settlements

Periodic Merkle batches that anchor groups of receipts into a single cryptographic proof. Each settlement shows the period window, receipt count, Merkle root, signature, and RFC 3161 timestamp.

Seal a new settlement using the button at the top. This takes all unsettled receipts, builds a Merkle tree over their payload hashes, signs the root, and requests an RFC 3161 timestamp from a trusted authority.

Click a settlement to see the sealed receipts listed individually, each linking back to its action. The detail page also includes a Python code snippet to verify any single receipt's Merkle inclusion proof offline.

Policies

Approval policies that govern what agents can do. Each policy has a name, mode, priority, and status.

Four policy modes:

Rules — deterministic conditions (e.g., "deny if amount > 50,000")
AI — plain-English policy evaluated by an LLM per action
Consensus — multiple models evaluate, high disagreement triggers review
Content Scan — regex patterns for PII, credentials, prompt injection

Policies are evaluated in priority order. The first matching policy's decision wins. Create, edit, or deactivate policies from this page.

Evidence & Compliance

Evidence

Sealed evidence packages bundling multiple actions into a single exportable artifact. Each package has a title, action count, status, and creation date.

Create a package by selecting actions from the list and providing a title. The package seals the selected actions with their receipts into an immutable, downloadable bundle.

Compliance

Compliance bundles mapped to regulatory frameworks. Choose a framework (EU AI Act Article 12, ISO 42001, AIUC-1, SOC 2), select a time period, and seal a Merkle-rooted evidence export.

The bundle includes every receipt in the period, organized by the framework's control requirements, with a Merkle root and signature for tamper-evidence.

Estate

Agent lifecycle management and compliance snapshots.

Compliance snapshots are point-in-time attestations of an agent's compliance status against a framework. Each snapshot shows pass/fail/pending findings per requirement. Create snapshots regularly — the dashboard warns if active agents have no recent snapshots.

Agent wills and decommission certificates are managed from the agent detail page, but the Estate section gives a cross-agent view of lifecycle status.

Escrow

Liability ledger for high-stakes agent operations. Create escrow accounts, record deposits before risky actions, and release or dispute after outcomes.

Escrow is available on Business and Enterprise plans. Free and Pro users see an upgrade prompt.

Configuration

Models

Three tabs:

Registered Models — custom model endpoints you've registered (BYOM). Each model has an endpoint URL, provider, and test status.
Providers — API key configuration for Anthropic, OpenAI, Google, AWS Bedrock, and other providers. Keys are encrypted at rest.
Preferences — toggle which models are available in Ask Aira chat, Cases, and AI policies. Set the default chat model here.

API Keys

Create and manage API keys for SDK access. Each key shows its prefix, name, creation date, and last used date.

Create a key, copy the full value immediately (it's only shown once), and use it in your SDK: Aira(api_key="aira_live_...").

Revoke a key to disable it permanently. Revoked keys can't be un-revoked — create a new one.

Webhooks

Subscribe to events like case.complete, action.notarized, case.requires_human_review, and others. Each webhook has a URL, event list, and delivery status.

Test a webhook to send a sample payload and verify your endpoint responds with 200.

Endpoints

Configure custom API endpoints for private deployments. Most users don't need this — it's for self-hosted setups where the API runs on your infrastructure.

Organization

Team

Invite team members by email with a role:

Owner — full access, can transfer ownership or delete the org
Admin — full access except ownership transfer and org deletion
Member — can use all features but can't manage team, billing, or settings
Viewer — read-only access to all data

Pending invites appear above the member list and can be cancelled.

Register at airaproof.com/register
Create an API key at Dashboard > API Keys
Register your first agent at Dashboard > Agents
Create a policy at Dashboard > Policies (optional — actions are allowed by default)
Run your first authorize + notarize via the SDK or API

Investigate an action

Go to Actions, find the action by type/agent/date
Click through to the detail page
See the policy evaluation (which policy matched, what it decided, reasoning)
See the receipt (Ed25519 signature, payload hash, signed payload)
Optionally set a legal hold if the action is under investigation

Prepare for an audit

Go to Compliance, seal a new bundle for the relevant framework + time period
Go to Estate, create compliance snapshots for each agent
Go to Settlements, seal a new settlement to anchor all recent receipts
Export the compliance bundle — hand the JSON file to the auditor
Point the auditor to the public verification URL: api.airaproof.com/api/v1/verify/action/{id}

Monitor agent behavior

Go to Agents > [agent] > Drift to see baseline vs current behavior
Ask Aira: "is any agent drifting?" for a cross-agent check
Set up a Webhook for drift.alert to get notified when divergence spikes