The platform

One infrastructure. Three jobs.

Sensitive data and models stay off-chain. Optional blockchain checkpoints are used only when another party must verify the same approved version.

01 · Automate

Workflow Control

Calculations, monitoring, and official reports run with identity, policy, and approval built in — not reconstructed after the fact.

02 · Govern AI

AI Assurance

An independent control boundary: the model may investigate and request tools; it does not become the source of authority, execution trust, or audit evidence.

03 · Tokenization-ready

Verified Records

Loan, collateral, and cashflow state with provenance. We prepare operating data; we do not issue tokens, custody assets, or replace legal work.

Agent may

Reason

  • Interpret the request
  • Retrieve scoped context
  • Draft and recommend
  • Request approved tools
OpenBITS checks

Control

  • Workload identity and purpose
  • Bank rules: data, tools, limits
  • Exact approval binding
  • Runner and signed receipt
Institution owns

Authority

  • Source-of-truth systems
  • Official books and records
  • Approver authority
  • Settlement and legal finality
Signed evidence

What the bank can show when the run finishes

A reconstructable pack — sources, model version, policy outcome, approvals, and output commitment — not a chat log.

evidence_pack.json 
// Bound to the approved daily risk report
{
  "workflow": "DAILY_PORTFOLIO_RISK",
  "proposed_by": "agent:risk-ops",
  "authorized_context": [
    { "type": "market_snapshot", "commitment": "0x9f2c…" },
    { "type": "position_file", "commitment": "0x41ab…" }
  ],
  "model_version": "risk_engine_v3.2",
  "policy_outcome": "PASS",
  "human_approvals": [
    { "role": "RiskApprover", "sig": "0x…" }
  ],
  "output_commitment": "0xab12…"
}

Architecture notes and peer-reviewed foundations are available to institutional reviewers. Talk to OpenBITS · Insights

Talk to OpenBITS