ONIVA
Get started
Oniva Agents · Available now

Agents your complianceteam can sign off on.

Build, run, and govern AI agent workflows across your existing systems. Every step is checkpointed, supervised by people, and signed.

Get startedView on GitHub

Source-available under BSL 1.1. Works with any LLM provider.

A run, up close

Watch the work, then approve it.

A KYC agent plans its steps, calls tools, and checkpoints as it goes. Before it creates a case, it stops and asks.

KYC agent working

  1. Pull customer record
  2. Screen against OFAC (Pending)
  3. Draft case file (Pending)
  4. Reviewer approval (Pending)
Run log
[run 0412] start workflow kyc-review (tenant: acme-bank)
✓ state: PLANNING -> TOOL_SELECTION
✓ core_banking.get_customer 200 184ms
✓ ofac.screen no match 310ms
! masked tool: payments.release_hold (state: REVIEW)
✓ checkpoint saved step 6/9
[run 0412] awaiting approval: servicenow.create_case
Open a case for the compliance reviewer with the drafted KYC file attached.servicenow.create_case
Approval required
{
  "customer": "C-20418",
  "risk": "medium",
  "attachments": 2
}
Case INC0041877 created

Product demo

See it end to end.

Oniva agent studio with built-in Compliance, KYC/AML, Market Abuse, Analysis and Data Retrieval agents
Agent studio: built-in Compliance, KYC/AML, Market Abuse, Analysis, and Data Retrieval agents, plus your own.
Oniva connectors for ServiceNow, Slack, Salesforce, Jira and GitHub
Connectors: ServiceNow, Slack, Salesforce, Jira, GitHub, and any OpenAPI spec.

Platform

From prompt to a decision you can defend.

Visual workflow builder

A canvas with 30+ node types: agents, logic, transforms, and integrations. Execution streams onto the canvas as it runs.

Durable execution

Every step is checkpointed. Crashes and restarts lose nothing, and runs of 8+ hours resume where they stopped.

State-machine safety

Agents run inside a 7-state machine. Tools that are invalid in the current state are hidden from the model, not rejected after the fact.

Signed audit log

A hash-chained decision log signed with RSA-2048. Replayable, kept for 7 years, and built for SOX, SEC, and MiFID II-style review.

Connectors

Native MCP servers for ServiceNow, Microsoft Teams, and OFAC, plus an OpenAPI factory. Point it at any spec and get a sandboxed tool with SSRF protection.

Tenant isolation

Postgres row-level security keeps every tenant's workflows, runs, and data separate at the database, not just in application code.

Human in the loop

Approval queues for high-risk steps. A reviewer sees the exact tool call and arguments before anything runs.

Three-tier memory

Agents keep working, session, and long-term memory, so context carries across steps and runs without bloating every prompt.

Test and deploy gates

Workflows are tested before they ship and promoted through deploy gates, so a change to a live flow is a decision, not an accident.

30+

node types on the canvas

7

states in the safety machine

35+

connectors, native and OpenAPI-derived

7 yr

audit log retention

Use cases

Built for operations, risk, and compliance.

Pull customer data from core systems, screen against OFAC, and draft a reviewer-ready case file. Every step is logged and signed, and a compliance analyst approves before anything is filed.

FAQ

Common questions.

Source-available under the Business Source License 1.1. Converts to Apache 2.0 in 2030. Commercial license on request.

Get started with Oniva Agents.

Leave your work email. We will help you set up, from one pilot workflow to production.