Back to vendors
R

Rivvun AI

Also known as: Rivvun, Rivvun AI Inc.

Visit site
Entry priceContact for pricingFull pricing detail

Autonomous execution layer for enterprise P&L integrity from Icertis veterans: nine named Steward and Sentinel agents verify spend pre-commitment and enforce revenue terms deal-to-cash, on an ontology-backed knowledge graph, YAML playbook DAGs, an MCP-native connector mesh, customer-defined approval gates, and stated SOC 2 Type I.

Rivvun AI Inc. is a Seattle company, with offices in Singapore and Pune, building an autonomous execution layer for enterprise finance that detects revenue and spend leakage in real time and runs governed corrective actions inside a company's own systems. Founded in 2026 by former Icertis executives Anand Veerkar and Niranjan Umarane with Patrick Linton, and backed by a $7.55 million oversubscribed seed co-led by Sitara Capital and 3one4 Capital, it ships nine named agents in two families.

The Stewards on the buy side, Invoice, Spend, Leakage, Margin and Supplier, verify spend against entitlements and operational evidence before it commits, and the Sentinels on the sell side, Renewal, Revenue, Customer and Margin Bridge, enforce pricing, milestones and commitments from deal close to cash collection. The platform's Commercial Data Foundation normalizes fragmented enterprise data into an ontology-backed knowledge graph preserving lineage, confidence and evidence.

Declarative YAML playbooks convert SOPs into versioned execution DAGs run by a three-tier agent hierarchy with complexity-based model routing, and an MCP-native connector mesh reaches contracts, CRM, ERP and procurement systems through governed, credential-free access, with a bi-directional API and event surface and plugin-based extensibility. Oversight is customer-defined.

Approval packs with auto-assembled context, stateful pause-review-resume, and policy-as-code guardrails determine what runs autonomously, with agent reasoning streamed in real time and every workflow producing a structured record of actions, approvals and P&L impact. Domain playbooks ship as plugin-isolated vertical units for lifesciences and healthcare, banking, professional services and industrial markets, and production case studies span an $11.2 billion satellite communications supplier consolidation, real-time royalty verification for a global publisher, and regulatory remediation at a major bank. The company reports SOC 2 Type I attestation with Type II underway, enterprise SSO and customer-controlled keys. Pricing is unpublished, with a Value Scan as the entry route.

Vendor details

Canonical URL

https://rivvun.ai

Category

Enterprise operations agent

Subcategory

Finance revenue and spend orchestration

Funding status

Independent, founded in 2026 by Anand Veerkar and Niranjan Umarane, former senior executives at Icertis, alongside serial entrepreneur Patrick Linton. Raised a 7.55 million dollar oversubscribed seed round co led by Sitara Capital and 3one4 Capital. Headquartered in Seattle with an engineering center in Pune, India.

Company status

independent

Use cases & customers

Primary use cases

spend assurance and rebate recoverymargin and revenue leakage defensecontract obligation enforcementtransaction level financial recovery

Target customers

enterprisemid-market

Deployment options

SaaS

Integrations

Installs as a non disruptive overlay on existing ERP, CRM, and procurement systems, reading transaction and commercial obligation data and executing governed corrective actions inside those source systems. No new system of record and no rip and replace, with a value scan against customer data before deployment.

In practice

Your procurement team negotiated rebates and volume commitments that never fully land. Rivvun's Steward agents verify each dollar against entitlements at the point of decision and recover what would otherwise leak to write offs.

Revenue leaves the P&L between deal close and cash collection through pricing and settlement gaps. Rivvun's Sentinel agents enforce commitments and pricing integrity across the sell side with an audit ready trail.

A regulated finance team cannot let AI act unsupervised on suppliers. Rivvun runs within policy guardrails, executing autonomously where allowed and routing higher risk recoveries to human review.

Agentic Index coverage score

9.0 / 14 capabilities · 64%

Integrations & Tool Calling Partial

Rivvun describes its integration architecture in depth. An MCP-native tool mesh governs all enterprise integration, with pre-built connectors for contracts, CRM, ERP and procurement, plugin-based extensibility where a new integration is a manifest, and governed data access through policy, not credentials. No target system is named, whether an ERP vendor, a CRM or a procurement platform, so Rivvun does not say which systems the connectors support.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Workflow Orchestration Full

Customers compose the flow, not just run a pipeline. Declarative YAML playbooks convert SOPs into versioned execution DAGs with dependency ordering, input contracts and failure policies. A three-tier hierarchy has a supervisor planning, task agents reasoning with tools and sub-workflows executing deterministically. Conditional DAG edges branch on real-world outcomes with state checkpointed at every node, and playbooks compile directly to execution plans. Customers encode their own SOPs into the playbook layer.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Knowledge Grounding & RAG Full

The Commercial Data Foundation is a knowledge graph Rivvun maintains. It maps source data into an ontology-first model through a semantic normalization pipeline, extends it with industry-specific objects and rules, and preserves lineage, confidence, evidence and policy-aware relationships in a knowledge graph with trust controls, the corpus every valuation and recovery action draws on. The consolidated supplier graphs in the production case studies are this layer at work.

SourceRivvun AI, rivvun.ai/technical-architecture and rivvun.airead 2026-09-07

Human Oversight & Guardrails Full

The gate is Rivvun's own, with the scope set by the customer. Approval packs and decision prompts arrive with full context assembled automatically, and stateful human-in-the-loop execution pauses, gets reviewed and resumes with no loss of fidelity. Guardrails are policy-as-code, declarative, versioned and rollback-ready, and enterprises define what runs autonomously and what requires human review. In the bank case study, remediation ran within the customer's policy guardrails.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Security, Identity & Governance Full

Rivvun holds a SOC 2 Type I attestation, described as audited controls in hand, with Type II marked coming soon. The controls customers get are enterprise SSO, customer-controlled keys, tenant isolation and zero-trust execution, where agents hold no credentials and every capability is scoped and enforced at runtime under versioned, rollback-ready policy-as-code. GDPR ready and HIPAA ready are suitability claims, not certifications.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Observability & Auditability Full

The audit trail is the execution log. Agent reasoning streams in real time with intermediate steps and evidence, every workflow generates a structured record of P&L impact, actions, approvals and evidence in outcome dashboards, and the knowledge graph preserves lineage and evidence beneath it. The LangGraph checkpointing Rivvun names is a build dependency, and the surfaces customers see are the reasoning stream, the outcome records and the lineage.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Memory & State Persistence Partial

Run state persists as a platform property. State is checkpointed at every DAG node, and stateful human-in-the-loop execution pauses, gets reviewed and resumes with no loss of fidelity, durable execution the customer relies on mid-workflow. There is no memory beyond a run, with no cross-session store, scope, lifetime or deletion path.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Deployment & Data Residency Not documented

There is no deployment option, no hosting region, no customer environment, and no self-hosted or private topology. The language about working inside the enterprise perimeter describes where corrective actions execute, in the customer's source systems, not where the platform is hosted.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Prebuilt Agents, Templates & Packs Full

Nine named agents in two families make up a browsable catalog. The Stewards are Invoice, Spend, Leakage, Margin and Supplier, and the Sentinels are Renewal, Revenue, Customer and Margin Bridge, each with its own place on the agents pages. Domain playbooks for healthcare compliance, invoice validation and KYC onboarding ship independently as plugin-isolated units. Plugin isolation is the design, and the production case studies show selective deployment.

SourceRivvun AI, rivvun.ai/agents and rivvun.ai/technical-architectureread 2026-09-07

Triggers & Channel Coverage Full

Customers can use the trigger surface directly. A bi-directional API and event surface triggers agents from anywhere and streams outcomes everywhere, transactions fire assessment at the point of decision before spend commits, the Sentinels enforce renewals, milestones and escalations as dated events, and delivery embeds in Slack, Teams, email and ERP portals.

SourceRivvun AI, rivvun.ai/technical-architecture and rivvun.airead 2026-09-07

Model Flexibility & Routing Partial

Rivvun routes across models on its own side, using complexity-based model routing that selects right-sized AI per task across tiers, balancing cost, latency and capability. Rivvun controls the routing, and customers and admins cannot select a model.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

APIs, SDKs & MCP Extensibility Partial

A bi-directional API and event surface triggers agents from anywhere, and plugin-based extensibility makes a new integration a manifest, so the platform is callable from outside by Rivvun's own description. There is no public API reference with authentication and endpoints. The MCP-native mesh is the agents' outbound tooling into enterprise systems.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

Testing, Debugging & Optimization Not documented

There is no evaluation surface for the agents. The Value Scan run against customer data before deployment is Rivvun's pre-sales practice, and schema validation at load time with semantic versioning on playbooks is build-time change governance with no readable comparable result. There is no harness, test set, score or release gate on the agents for customers.

SourceRivvun AI, rivvun.ai/technical-architecture and rivvun.airead 2026-09-07

Browser & Computer Use Not documented

The agents act inside enterprise source systems through the MCP-native connector mesh, not by driving a browser or operating a computer interface, and there is no interface control of any kind.

SourceRivvun AI, rivvun.ai/technical-architectureread 2026-09-07

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Contact for pricing

not disclosed

Trial available

What is public

No public rate. Rivvun publishes no list pricing; it offers a Value Scan request and an executive briefing, and routes commercial terms through sales.

Billing mechanics

Not disclosed. Engagements begin with a Value Scan against customer data that estimates recoverable value, followed by a proof of value; no billing unit or rate is published.

Cost watchouts

Because pricing is not disclosed and may combine a platform fee with a recovery share, confirm the split, the attribution method for recovered value, and integration effort across ERP, CRM, and procurement systems.

Variable cost rationale

If priced on recovered value the cost is largely self funding and scales with recoveries, but the exact model is not public and may include a platform fee, so predictability depends on the undisclosed structure.

Additional watchouts

The billing model is not public; ask whether pricing is a platform fee, a share of recovered value or both, and how recovered value is measured and attributed.

Sales call required

Yes, required for paid access

Free / trial

Free value scan estimating recoverable value with no upfront integration; proof of value engagement before contract

Key ambiguities

No entry rate or billing model is published; commercial terms are set through sales after a Value Scan request or an executive briefing.

Agentic Index verified 2026-09-07

Alternatives to Rivvun AI

The closest documented capability profiles to Rivvun AI among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

  • Abacum9.0 / 14Matches Rivvun AI across all 14 documented capabilities
  • Campfire9.0 / 14Fuller documented coverage on APIs, SDKs & MCP Extensibility
  • Cube10.0 / 14Adds documented Testing, Debugging & Optimization
  • Byron8.5 / 14Fuller documented coverage on Integrations & Tool Calling
  • Fieldguide7.5 / 14A lighter documented profile than Rivvun AI
  • Ironclad9.5 / 14Fuller documented coverage on Integrations & Tool Calling and APIs, SDKs & MCP Extensibility

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.