Foundation · App 00
Let AI agents do real work without losing control
One identity, policy, approval, audit, metering and channel layer for people and agents alike. Included in every One3 subscription and being built from Wave 0.
One governed foundation where every human, AI agent, integration and channel acts through the same Action contract, inside a tenant boundary that a company, an agency or a regulator can trust. The Platform is not a product customers shop for. It is the reason the business apps feel like one system: one login, one admin console, one approvals inbox, one audit trail, one bill, one Copilot, one set of connectors and one place where AI cost and AI risk are controlled.
What the Platform is
The Platform is included in every One3 subscription. Customers do not buy it separately; limits and usage vary by plan. It turns the One3 Kernel and control plane into 24 modules that tenant admins, agency partners and One3 operators use every day: Tenants & Workspaces, Identity & Access, Admin & Policy, the Kernel Action Bus, Notifications, Trust & Audit, Metering & Subscriptions, the Approvals Inbox, Copilot, Model Hub, AI Observability, Integration Hub, MCP Hub, Migration Factory, Connector Health Center, the Fact Register, and the WhatsApp, voice, email, meeting, browser and desktop channels.
Three commitments
- Every write is an Action. The UI, the REST API, an MCP client, a workflow, Copilot, a rented Digital Employee, a WhatsApp message or a voice call all invoke the same typed Action through the Kernel Action Bus. Each call gets the same entitlement, permission, policy, risk-tier, budget, approval, audit and cost handling.
- Autonomy is earned and provable. Risk tiers R0 to R4, reversibility classes, delegation-of-authority limits and the Approvals Inbox are designed to make it safe to let agents act. Trust & Audit and AI Observability make it provable afterwards.
- Adoption is engineered, not assumed. Migration Factory, Connector Health Center and the Fact Register exist because switching risk, stale data and changing rules stop real companies from adopting software.
How it works with the rest of One3
Every business app, from CRM & Sales to Finance, depends on the Platform from Wave 0. A deal update in CRM, a campaign send in Marketing and an invoice in Billing all run through the same Action Bus. That means one audit trail across apps, one approvals inbox for every risky step, and one usage meter for AI credits and messages. Agents hired in the AI Workforce app get identities whose permissions are always a subset of their human owner's. Channels such as WhatsApp and voice are interfaces, not products: they reach every subscribed module through the same Action contract and carry usage meters rather than app subscriptions.
Built for agencies and Indian SMBs first
The Year-1 focus is digital agencies managing client sub-accounts and owner-led growth teams. For them the plan includes a partner tier with white-label domains, sub-account snapshots, pooled allocation and rebilling reports, WhatsApp approvals for owners who are away from a desk, GST fields on One3 invoices (to be verified), and an India region by default. Mid-market, regulated and enterprise needs such as SCIM, SIEM streaming and silo tenancy are planned for later waves.
Problems solved
What Platform fixes
Too many logins and admin consoles
Staff juggle separate logins, roles and admin consoles per tool, and leavers keep access. The Platform plans one global login with per-tenant memberships, SSO, roles generated from app manifests and joiner/leaver rules.
Unclear AI agent access
AI agents act with human-level or unclear access and nobody can say who approved what. Agent identities get permissions that are a subset of their owner's, with R0–R4 risk tiers, a delegation-of-authority matrix, approvals with preview and a hash-chained audit trail.
Unpredictable AI spend
AI spend is invisible per team, client or outcome. AI credits, pass-through meters, budgets with auto-pause and a cost record on every call are designed to make it visible and capped.
Manual agency rebilling
Agencies run dozens of client accounts across tools with manual rebilling. A partner to tenant to workspace hierarchy, white-label domains, pool allocation, client price lists and rebilling reports address this.
Risky switching
Moving suites risks data loss, broken relations and lost consent. Migration Factory plans mapping, dry runs, batch checkpoints, reconciliation, rollback and consent carry-over.
Integrations that fail silently
Agents act on stale data when connectors break. Connector Health scores each connection 0–100, applies freshness targets and restricts agent autonomy in degraded mode.
Owners live on WhatsApp
Indian SMB owners run their business on WhatsApp and phone, not dashboards. WhatsApp approvals and owner commands, and later Indic voice, run over the same Actions.
Model lock-in
Choosing one model vendor locks in quality, cost and residency. Model Hub plans a registry, eval-driven routing, in-region routing for restricted data and bring-your-own provider keys.
Key capabilities
What you can do
Kernel Action Bus
One typed entry point for every state change, with entitlement, permission, policy, budget, approval, audit and cost checks on every call.
Approvals Inbox
One inbox for every decision routed to a human, with preview, risk, AI rationale and step-up. Approve from the web app, PWA or WhatsApp.
Hash-chained audit
Append-only audit events per tenant, chain verification, access logs for sensitive reads and data-principal request workflows.
Agency partner tier
Partner, tenant and workspace hierarchy with sub-account snapshots, white-label domains, pooled allocation and client price lists.
Copilot that acts through Actions
A chat panel and command bar that answers with citations and acts only through Actions the user is permitted to invoke.
Model Hub
The only path to any model: routing by task class and data classification, fallbacks, caching, bring-your-own keys and a cost row per call.
Migration Factory
Profile, map, dry-run, migrate in batches, reconcile and roll back by migration id. HubSpot first, then GoHighLevel and Zoho CRM.
Connector Health Center
A 0–100 health score per connection with freshness targets. Agents check health before acting on connector data.
WhatsApp as a governed channel
Templates, conversations, owner commands and approvals with consent checks, quality guards and pass-through metering.
Modules
24 modules, one data model
Core modules come with the app. Add-ons switch on when you need them. Some modules can be bought on their own.
Tenants & Workspaces
Partner, tenant and workspace hierarchy with white-label and sub-account snapshots
ExploreIdentity & Access
One login across tenants, agent identities, SSO and least-privilege access
ExploreAdmin & Policy
One admin console and one policy layer that every app, agent and channel obeys
ExploreKernel Action Bus
The single governed entry point for every state change and side effect
ExploreNotifications
One notification centre for every app, agent and platform event
ExploreTrust & Audit
Prove what happened, who did it and under which authority
ExploreMetering & Subscriptions
Entitlements, usage meters, budgets, invoices and agency rebilling in one place
ExploreApprovals Inbox
One inbox for every decision the Action Bus routes to a human
ExploreCopilot
One assistant across your apps that answers with citations and acts through Actions
ExploreModel Hub
The only path to any model, routed by task, data class, quality and cost
ExploreAI Observability
See what every AI call and agent run did, what it cost and where it failed
ExploreIntegration Hub
Certified connectors with checkpoints, replay, signed webhooks and governed writes
ExploreMCP Hub
Expose your One3 Actions to external AI assistants and consume approved MCP tools
ExploreMigration Factory
Switch to One3 with profiling, AI mapping, dry runs, reconciliation and rollback
ExploreConnector Health Center
A health score for every connection, and agents that stop acting on stale data
ExploreFact Register
Owned, expiring facts that block launches and decisions when they go stale
ExploreWhatsApp channel
WhatsApp as a governed channel to every One3 app, with consent and quality guards
ExploreLocalization
UI, templates and content in Indian languages, with glossaries and human review
ExploreVoice channel
Run One3 by phone in Indian languages, with disclosure, consent and strict risk limits
ExploreEmail-native agent
CC an agent mailbox to log emails, create tasks and draft replies for approval
ExploreMeeting companion
Capture in-person meetings with consent and turn them into tasks and record updates
ExploreBrowser & computer-use agent
Operate portals with no API in a sandbox, with screenshots and an approval gate
ExploreDesktop assistant
Bring Copilot to any desktop app with a hotkey and explicit, revocable grants
ExploreGenerative UI
Role- and intent-specific screens generated from certified components
ExploreWorkflows
How the work flows
End-to-end workflows inside the app and across One3.
Agency client onboarding
An agency partner creates a new client sub-account from a snapshot, allocates seats and credits, and brands it, all in one flow.
- 1.Partner admin opens the agency console and chooses New sub-account.
- 2.Selects a snapshot or edition that carries apps, pipelines, templates, automations, roles and notification defaults.
- 3.The tenant provisioner creates the tenant, default workspace and owner membership.
- 4.Partner allocates seats and AI credits from the agency pool and sets the client price list.
- 5.White-label domain and sender name are applied; DNS and TLS are verified.
- + 1 more steps →
Agent Action with human approval
A Digital Employee proposes a risky Action; the Action Bus pauses it and routes it to the right approver with a clear preview.
- 1.An agent invokes a typed Action such as sending a WhatsApp campaign message.
- 2.The Action Bus checks entitlement, permission, policy, delegation of authority, budget and connector health.
- 3.Because the Action is risk tier R3, an approval request is created with preview, risk, reversibility warning and AI rationale.
- 4.The approver resolved from the DoA matrix approves, rejects, edits or asks for more information, from the web, PWA or WhatsApp.
- 5.The durable workflow resumes, executes in a transaction with the outbox and records audit, cost and trace.
Switch from an incumbent tool
Migration Factory moves records from HubSpot, GoHighLevel, Zoho CRM or spreadsheets with a dry run, reconciliation and rollback.
- 1.Connect the source and run profiling to count objects, custom fields, owners and pipelines.
- 2.Review AI-suggested mappings with confidence and reasons in Mapping Studio, then approve them.
- 3.Run a dry run with full validation and no writes; fix errors by type.
- 4.Execute in resumable batches that write through the target module's Actions.
- 5.Review the reconciliation report with counts, sums and a sample diff, and sign off.
- + 1 more steps →
Data-principal request
A person asks to access, correct or erase their data; Trust & Audit runs the request across apps with evidence.
- 1.The request arrives through the portal or a staff member logs it.
- 2.Identity is verified and the person is resolved across apps.
- 3.The DPDP request assistant drafts a data location report and an erasure plan with blockers such as statutory retention or legal hold.
- 4.The DPO reviews and executes the erasure with step-up authentication.
- 5.Evidence is stored and a completion notice is sent.
Owner approves from WhatsApp
An SMB owner away from a desk reviews and decides on a pending approval inside WhatsApp.
- 1.A discount above the owner's DoA threshold creates an approval request.
- 2.The owner receives a WhatsApp utility notification bound to their verified identity.
- 3.The owner reviews the summary and approves or rejects with a reason.
- 4.The decision is recorded in audit with the channel, and the underlying workflow resumes.
AI inside
AI skills that act through governed Actions
Every AI skill runs through the same typed Actions people use, with risk tiers, approvals and an audit trail.
Onboarding agent A1
Proposes apps, modules, workspace structure and roles from a business description. The admin applies each step with normal Actions; there is no apply-everything button by design.
Access-review agent A2
Flags unused permissions, dormant accounts, over-privileged agents and SoD risks as proposals. Revocations are always human-approved.
Plain-language policy drafting A1
Turns a rule written in plain language into a typed rule with test cases. An admin activates it after simulation.
Approval summaries
Explains what changes, who is affected, risk factors and similar past decisions with citations, shown beside the raw preview and labelled as AI.
Audit Q&A
Answers questions about the audit trail with citations to event ids. Read-only and permission-filtered.
DPDP request assistant A2
Drafts party resolution, data location reports and erasure plans with blockers. Erasure stays a DPO decision with step-up.
Field-mapping and sync-error fixer A2
Suggests connector and migration mappings with confidence and reasons, and proposes fixes for sync errors. Never auto-applied.
Failure clustering and drift alerts A3
Groups AI failures weekly by error signature and intent, and flags score, cost and latency regressions per route.
Owner commands over WhatsApp
Lets owners ask Copilot and trigger allowed Actions from WhatsApp; R3 Actions go to the approval flow.
Digital Employees
Agents that operate Platform
Rent them by the month. They work under your team's supervision and their cost and returns appear in the Investment & ROI Ledger.
AI Agent Governance Auditor
A meta-role that samples traces and the action log, scans for policy compliance and drafts drift reports and AI incidents. Skill demotion stays a proposal.
AI Access Reviewer
Prepares access review campaigns, reports dormant accounts and proposes revocations that a human approves.
Learn moreAI Privacy (DPDP) Operations Assistant
Triages data-principal requests, prepares verification checklists, party resolution reports and erasure plan drafts. Advisory; the DPO decides.
Learn moreAI Policy Writer
Drafts policy rules with tests and Company Constitution sections. Activation is a human Action after simulation.
Learn moreAI Data Engineer
Proposes connector mappings and sync-error fixes and requests dead-letter redrives through approval.
Learn moreAI Data Quality Steward
Proposes dedupes during migrations and triages quarantined records. Merges stay human.
Learn moreAI Meeting Scribe
Produces transcripts, summaries and action items from consented meetings; tasks are created only after confirmation.
Learn moreAI Chief of Staff
Sends daily briefings and approvals digests over WhatsApp and, later, voice.
Learn moreIntegrations
Connects with the tools you already use
- Google sign-in
- Microsoft sign-in
- Microsoft Entra ID
- Okta
- Google Workspace
- Slack
- Microsoft Teams
- WhatsApp Business Platform
- Gmail
- Google Sheets
- HubSpot
- HighLevel (GoHighLevel)
- Zoho CRM
- Pipedrive
- Freshdesk
- Payment gateway (cards, UPI AutoPay)
- GSP / IRP e-invoicing
- LiteLLM
- Langfuse
- Generic webhooks
- MCP clients and servers
Connector availability depends on provider access and is verified before launch. How integrations work →
Security & compliance
Built to be trusted
Tenant isolation by design
Row-level security is forced on every tenant table, cache keys are tenant-prefixed, and MCP endpoints, webhooks and notifications are tenant-bound. The release gate is zero cross-tenant leaks across UI, API, MCP, Copilot, agents, webhooks and notifications.
Layered authorisation
Every request is checked for entitlement, licence, role permission, scope and record sharing, field-level security and conditions such as risk tier, DoA, budget, consent and step-up. Hiding a button in the UI is never the control.
Evidence for every Action
Every Action is written to the action log, audit events are hash-chained with a verification job, and reads of sensitive fields and exports are logged.
Agents never approve
An agent cannot hold an R4 approver permission or approve what it proposed. Requester and approver must differ. Untrusted content such as emails or web pages cannot trigger R3 or R4 Actions.
India region by default
Pooled tenants are planned to run in an India cell, with restricted data routed to in-region or local models. DPDP, CERT-In and sector obligations are pointers only; verify with counsel.
Tenant data does not train shared models
By default, tenant data is not used to train shared models. Safety controls such as injection detection, PII masking and residency routing cannot be turned off by a tenant.
Industry Editions
Industries that use Platform
Frequently asked questions
Do I pay separately for the Platform?
No. The Platform is included in every One3 subscription. Limits vary by plan, and some items carry their own line: premium connectors, assisted migration as a service, and usage meters such as AI credits, messages and voice minutes. No prices are set in the plan yet.
Is the Platform available today?
It is being built in waves. The core foundation is planned for Wave 0 and Wave 1, with the agency partner tier, WhatsApp channel GA and SSO in Wave 2. Voice, localization tooling and the email-native agent are planned for Wave 3, and the browser agent, desktop assistant and Generative UI for Wave 4.
How do AI agents stay under control?
Agents act only through typed Actions. Each Action has a risk tier from R0 to R4 and a reversibility class. Risky Actions pause for a human approver chosen by your delegation-of-authority rules, and R4 Actions always need a human. Everything is audited and costed.
Can I use my own AI provider keys?
Model Hub is planned to support bring-your-own provider keys on higher plans. Usage is still traced, and model cost is not charged to your AI credits where your own key applies.
Can I move from HubSpot, GoHighLevel or Zoho CRM?
Migration Factory is planned to start with CSV, XLSX, Google Sheets and HubSpot in Wave 1, then GoHighLevel and Zoho CRM in Wave 2, with dry runs, reconciliation and rollback. Provider access terms are recorded in the Fact Register and verified before build.
Where is my data stored?
Pooled tenants are planned to run in an India region by default. Bridge, silo and sovereign options are planned for later waves. Residency rules are tracked as facts to verify, not as legal advice.
Can external AI assistants use my One3 data?
Yes, through MCP Hub, planned for Wave 1. You grant an assistant scoped, revocable access with OAuth. It can only call Actions you are entitled to, and R3 or R4 Actions still need approval inside One3.
Works with
Related apps
CRM & Sales
A WhatsApp-first CRM where sellers and Digital Employees work the same pipeline
ExploreMarketing
Consented, measurable marketing across WhatsApp, email, search, ads and social
ExploreService Desk
Support on every channel, resolved on the same customer, invoice and order records
ExploreBuild
Automations, forms, surveys, documents and custom apps on the data you already run
ExploreIntelligence
Cited answers, one set of metrics and decisions tracked to a measured result
ExploreAI Workforce
Rent ready-made AI Digital Employees that work for your team, with a monthly ROI statement
ExplorePlan your Platform rollout with us
Tell us about your team and the outcomes you want. We will walk you through the apps, the Digital Employees and the plan for your industry.