8 AI Personas. Auto-Routed by Intent. Scoped by Role.
AeionClaw doesn't load all 47 skills + 89 MCP tools on every conversation — that would waste tokens, leak context, and bog down latency. Instead, 8 domain-specific personas auto-route based on user intent. Ask about an order? Commerce-Ops. Ticket question? Support-Agent. Invoice query? Finance-Manager. Each persona has scoped tool access, persona-specific system prompts, and memory partitioned to its domain.
How Persona Auto-Routing Works
The persona router runs before every conversation turn. It analyzes the user's message (or the triggering event), matches it against persona-keyword patterns + recent context + user role, and selects the persona with the highest confidence score.
The 8 Built-In Personas (seeded on first boot)
Each ships as a built-in persona record, auto-selected by context. Customize via the admin UI or override the prompt entirely. Configure new personas for your own domain.
general
Default fallback for cross-module questions, "how do I do X?", platform-navigation help. Read-only across all collections within RBAC scope.
commerce-ops
Orders, products, inventory, fulfillment. Composes with `aeion-commerce-ops`, `aeion-pricing-engine`, `aeion-fulfillment` skills. Memory: order + customer history.
crm-advisor
Deals, contacts, pipeline, follow-ups. Composes with `aeion-crm-assistant`, `aeion-marketing` skills. Memory: contact + deal history.
support-agent
Tickets, customer service, KB lookup, draft replies. Composes with `aeion-helpdesk-agent`, `aeion-inbox`, `aeion-connect` skills. Memory: ticket history + similar resolutions.
content-editor
SEO, content briefs, headline drafts, multilingual variants. Composes with `aeion-content-manager`, `aeion-marketing`, `aeion-aeionica` skills. Memory: brand voice + recent content history.
data-analyst
Cross-module analytics, cohort queries, narrative generation from metrics. Composes with `aeion-analytics-reporter`, `aeion-bi` skills. Memory: query history + dashboard definitions.
finance-controller
Invoices, expenses, financial reporting, variance commentary, board materials. Composes with `aeion-finance-manager`, `aeion-executive` skills. Memory: transaction history + period comparisons.
talent-recruiter
Resume screening, candidate-job matching, interview scheduling, offer-stage tracking. Composes with `aeion-recruit-assistant`, `aeion-lms` skills. Memory: candidate pipeline + role history.
Routing Decision Walkthrough
Real router decision for a sample user query.
Persona Wins Beyond Token Efficiency
Operational benefits of scoped personas.
Faster latency
smaller prompt = lower TTFT
Lower cost
fewer tokens per call
Better accuracy
focused context = less hallucination
RBAC scoping easier
persona has explicit tool allowlist
Audit clarity
every action attributed to a specific persona
Custom personas easy
add domain-specific personas without code
Memory partitioning
finance-controller's memory doesn't pollute support-agent
Approval workflows per persona
finance-controller requires CFO approval for refunds; support-agent doesn't
Building Custom Personas
Most customers configure 1-3 custom personas on top of the 8 built-ins.
Why scoped personas beat "one big AI assistant."
The "load all the tools into the prompt" pattern (popular early Anthropic + OpenAI Assistants) hits scaling walls fast — 89 MCP tools = ~50K tokens of tool schemas in every prompt. That blows latency and cost. Aeion's persona-routing pattern keeps each prompt focused (3-15 tools) which makes inference 10-20x cheaper and 3-5x faster. The accuracy improvement from focused context is also measurable — task-completion rate goes up because the LLM isn't distracted by irrelevant capabilities.
8 personas. Auto-routed. Token-efficient. Audit-ready.
Configure custom personas for your domain. Each scoped to relevant skills, tools, and memory. Each audit-logged. Each RBAC-aware.