[ NEURALARC_INTERNAL_—_NEURALCRM ]
Agency Operating System
What it is
CRM, ERP and client portal in one, with an AI agent that can operate it.
[ THE_PROBLEM ]
Why this existed
Agencies run on six disconnected tools and a shared inbox. Clients chase updates because there is nowhere to see them, and the team re-enters the same information into a CRM, a project tracker and an invoicing tool.
[ WHAT_WE_BUILT ]
What we built
Staff side: clients, contacts, a drag-and-drop kanban sales pipeline, projects, preview links, feedback, tickets, tasks, invoices and team roles. Client side: a portal scoped so a client sees only their own company — live project previews rendered in-app, an inline feedback panel, tickets with threaded replies, shared files and a project timeline. Convex's reactive queries mean every list, badge and notification updates live across all signed-in users with no polling and no query cache. On top sits an AI layer with its own tool surface over CRM actions, agent runs, an API-key system and a public API core — both unit-tested. Plus a command menu, notifications, activity history and a seeded demo dataset.
- Clients, contacts and a drag-and-drop kanban sales pipeline
- Projects, preview links, feedback, tickets, tasks and invoices
- Team roles and permissions
- Client portal scoped to a single company
- Live in-app project previews with an inline feedback panel
- Tickets with threaded replies, shared files and a project timeline
- Live-updating lists, badges and notifications with no polling
- AI tool layer over CRM actions with tracked agent runs
- API key system and a public API core, both unit-tested
- Command menu, activity history and a seeded demo dataset
[ HOW_IT_IS_USED ]
How a company uses it
An agency or professional-services firm runs delivery and client communication in one system, and gives clients a portal instead of a thread — while the AI layer handles the "update the pipeline, draft the invoice, summarise this account" work.
Built with
[ COMMON_QUESTIONS ]
Questions clients ask
Why build rather than use an off-the-shelf CRM?
The client portal is the reason. Generic CRMs treat clients as records, not as users with a login. Giving a client a scoped view of their own projects, previews, tickets and invoices is the part that changes the relationship, and it is the part that is hardest to bolt onto existing tooling.
What does "live-updating with no polling" actually mean?
Queries are reactive subscriptions rather than fetches. When a ticket status changes, every open client and staff view reflects it immediately without a refresh and without a polling loop hammering the server. On a collaborative tool this removes an entire category of stale-data confusion.
Can the AI layer be trusted with client records?
It has a defined tool surface over CRM actions, and agent runs are tracked so what it did is inspectable. The design principle is that the AI has the same capabilities as a team member with the same role — not database access.
[ RELATED_WORK ]
Similar builds
Financial media
Editorial & Research Operations Console
The back office for a financial news operation — research, production, publishing and audio.
Read case studyLive events
Event Ticketing Operations Console
Everything an events team does between selling a ticket and letting someone through the door.
Read case studyMobility
Ride Platform Safety & Disputes Console
The console for handling disputes, SOS alerts and KYC on a ride-sharing platform.
Read case studyIs this close to your problem?
Most engagements start with a version of something on this page. Tell us what is different about yours and we will tell you what it changes.
Start a conversation