Copilot
Talk to your CRM
An AI copilot with streaming responses, aware of what you have on screen, that runs real actions in the pipeline — create a deal, move a stage, schedule a task, suggest the next message.
LNData Product · Toudai CRM In development · MVP in production
The CRM you talk to through AI.
Toudai is a CRM where the conversation is the interface: instead of navigating screens and forms, you ask in natural language — "open a deal for this company", "show me the overdue follow-ups", "sum up my week" — and an AI copilot runs the action in your sales pipeline. It is LNData’s AI-first CRM, built for Brazilian SMEs, with the salespeople’s WhatsApp finally inside the CRM.
Role · Proprietary product · architecture, full-stack engineering, applied AI and infrastructure
Status · Multi-tenant SaaS · in development · MVP already in production
In short, for decision-makers
The problem
Established CRMs leave two gaps for the Brazilian SME: "real" AI is locked to the top plan and billed in dollars, out of reach for the average company; and WhatsApp — where selling actually happens — is a black box on the salesperson’s phone, with no visibility, no auditing and no history in the CRM.
What LNData is building
Toudai puts AI at the center, not as an add-on: you talk to the CRM and the copilot does the work for you. The salespeople’s WhatsApp is mirrored, audited hierarchically and turned into history inside the CRM. All multi-tenant, with AI included in every plan and pricing in local currency.
The promise
Less time filling in screens, more time selling: the sales operation fits into a conversation, with AI handling the record-keeping. Managers finally see what used to be invisible on WhatsApp, and the company gets a smart CRM without paying dollar prices for it.
AI-first
the conversation is the interface
SaaS
multi-tenant platform
WhatsApp
mirrored and audited
In beta
sellable MVP in progress
Natural language
ask, and the AI acts on the pipeline
MCP-native
one tool catalog for the AI
AI included
in every plan, metered by credits
Multi-tenant
schema isolation, LGPD by design
Architecture
You speak; the copilot interprets and picks the right tool from a single AI catalog; the CRM executes — always with human confirmation for writes. All orchestrated in containers, with slow tasks running in workers and responses streamed token by token.
You speak
- Natural-language chat
- Aware of screen context · what you are looking at
- Unified WhatsApp · real time
AI brain
Streaming copilot (SSE)
Internal MCP server
Per-tenant RAG (Qdrant)
LiteLLM · Claude / OpenAI
The CRM acts
- Creates a deal, moves a stage
- Schedules a task and follow-up
- "Apply" draft · you confirm
What the platform does
A full CRM underneath, a conversation on top. The whole sales core is there — the difference is that AI and WhatsApp are part of the product, not a plugin.
Trust
The AI proposes, you confirm
No write happens on its own: write tools return a draft that appears as an "Apply" card. The AI never writes to your CRM without your approval.
WhatsApp
WhatsApp inside the CRM
A unified real-time inbox of the salespeople’s WhatsApp conversations, with QR-code pairing, anti-ban (send spacing, daily caps) and media download.
Management
Hierarchical conversation auditing
The supervisor sees the team’s conversations, with a trail of who read what and a configurable rule for replying in another salesperson’s chat. Sales stops being a black box.
CRM core
A complete pipeline
Companies and contacts with dedup, leads with explicit conversion, deals with multiple pipelines and drag-and-drop kanban, a product catalog, activities, tasks and per-company custom fields.
Routine
AI daily briefing
The AI produces a daily summary of what matters — pending follow-ups, deals that went cold, the day’s priorities — so the salesperson starts knowing where to act.
The intelligence that makes the difference
What sets Toudai apart from a CRM with "a chatbot glued on" is where the AI lives: a single tool brain, with safe writes and knowledge isolated per company.
MCP-native architecture
A single internal MCP server powers the copilot with a tool catalog. The same "brain" will feed the messaging agent and the client’s external access — without reimplementing the intelligence three times.
Writes via draft, not direct execution
Write tools return action proposals instead of committing. The human confirms in the "Apply" card — the condition for trusting an AI that touches a real sales pipeline.
Per-company RAG and cost governance
Each company has its own vectorized, isolated knowledge base, and AI usage is metered per call — with quotas and credit-based billing, which makes including AI in every plan possible without blowing the cost.
Beyond the core
Multi-tenancy, RBAC and platform security
The conversation is the visible part; underneath is the platform infrastructure that sustains many companies at once, with isolation and security taken seriously — the foundation already complete and in production.
Physical multi-tenancy
One PostgreSQL schema per company, routed by subdomain, with async provisioning. Structural isolation for LGPD and clean offboarding, validated by a mandatory isolation suite in CI.
Scoped RBAC
Scoped permissions (own / team / all) over a multi-level team tree. Out-of-scope access returns 404 — it does not even reveal the resource exists.
Immutable auditing
A partitioned append-only trail with a transactional outbox pattern: a reliable record of everything that happened, with no external service.
Security taken seriously
Invite-code-only sign-up, 2FA TOTP, attempt lockout and Argon2 hashing. LGPD by design: the company is the controller, the platform the processor.
Djangodjango-tenantsPostgreSQLCelery
The vision
Operating the CRM becomes a conversation: you ask in natural language and the AI executes, taking the salesperson out of screen-filling.
WhatsApp — where selling actually happens — gains visibility, auditing and history inside the CRM.
Real AI included in every plan, metered by credits and priced in local currency — in the space established CRMs leave open for the SME.
Foundation complete and in production (multi-tenant, auth, RBAC, auditing); sellable MVP in progress, with features already live.
Platform stack
DjangoDRFReactCeleryPostgreSQLQdrantMCPLiteLLMDocker
Want to follow Toudai closely?
Toudai is in active development, with the MVP already in production. If your company wants a CRM you talk to through AI that brings WhatsApp inside, talk to LNData to follow the launch or join the early tests.