Cortex AI
Cross-app intelligence on your ledger and operations — chat, tools, and guardrails.
What is Cortex?
Cortex is Axiomatic’s built-in AI: a conversational interface to your business data and workflows. It runs in-product (dedicated Cortex area and chat entry points across modules) and is backed by tools that read and, where allowed, act on your tenant — always within identity, entity, and role boundaries.
Cortex is not a separate server or bolt-on SKU. It is part of the same application as Ledger, CRM, BI, and the rest, so answers can combine operational and financial context without exporting CSVs.
What you can do
- Ask in plain language — balances, trends, journal provenance, “what posted this?” style questions
- Classify and reconcile — imported transactions, source-aware suggestions, batch operations where enabled
- Work with documents — ingest, summarize, and map to accounts or projects where the product supports it
- Author and refine rules — help draft or explain DSL rules; execution still follows your approval and posting policies
- Navigate complexity — explain modules, reports, and workflows to onboard teammates
Exact tool coverage expands with releases; the in-app experience always shows what Cortex can touch for your role.
Guardrails
- Ledger authority — Posting, approvals, and irreversible actions follow the same permissions and credit/billing rules as the rest of the app. Cortex does not bypass them.
- Tenant isolation — Data access is scoped to your tenant and entity like any API route.
- Policy — Administrators can constrain AI behavior via Admin → AI policy (where available).
- Auditability — Actions that change financial state remain traceable through existing journal, event, and audit mechanisms.
Where to open Cortex
Use the Cortex area in the app for full conversations and history. Shortcut: G then C (when enabled). Module pages may expose contextual chat or assistants tied to that workflow.
Agent pods (Sales, PM, Design, Engineering)
Custom Cortex agents are entity-scoped sub-assistants with their own system prompt and toolPolicy (module scopes, read-only mode, optional tool allowlist).
Setup
- Open Cortex → Settings → Agents (
/cortex/settings/agents). - Create from a pod template (Sales SDR, Marketing, PM, Design, Engineering) or customize manually.
- Invoke in chat with
@agent:slugor sticky agent mode in the sidebar.
Cortex can also design agents via tools: describe_cortex_agent_schema, create_cortex_agent (through Approve / propose_action).
Product pod workflow
| Role | Agent | Key tools |
|---|---|---|
| PM | pm | add_task, update_project_task, list_project_tasks |
| Design | design | Documents (specs/wireframes), Website Studio (marketing site only), update_project_task for UX review notes |
| Engineering | eng-cursor | launch_cursor_agent — opens Cursor Cloud Agent PRs; never edits code in-app |
Project tasks support externalLinks: github_pr, cursor_run, spec_doc, vercel_preview (future). Link PRs manually with update_project_task or automatically via GitHub webhook (POST /api/webhooks/github/{entityId}) when the PR body or branch contains the task UUID (task/<uuid>/...).
Integrations
- GitHub — Admin → Integrations → Developer. Webhook secret verifies
X-Hub-Signature-256. - Cursor — API key +
defaultRepoon the connection; optional per-project override viajobs.metadata.githubRepo. Engineering agent callslaunch_cursor_agent.
Delegations (cron + granted permissions) and invocation history live on each agent’s detail page.
Learn more
- Tutorial: Using the AI Assistant — hands-on patterns, classification flows, and tips
- API Reference — programmatic automation alongside AI
- Platform overview — how Cortex fits the unified ERP
- Glossary — tenant, entity, credits, and other terms