Total AI Governance
Enter Statefold
Total AI Governance

Totalintelligence.Under control.

One operating system to see how AI is used, decide what is allowed, act where controls exist and prove the result. Every team works from the same evidence.

Enter Total Intelligence ↓ Explore the control plane ↓
01 Every perspective 02 Governed action 03 Verifiable efficacy

One intelligence layer across

Identity×Prompts×Agents×Tools×Models×Data×Policy×Risk×Cost×Evidence×
01 / The shared intelligence layer

One enterprise.
Six readings.
One record.

The same AI event can be a technology decision, a security incident, a privacy question, a risk exception, a financial signal and an executive concern. Where supported stages produce evidence, Statefold holds the complete chain; otherwise the gap stays explicit.

ONE SHARED RECORD Every perspective meets on the same fold. SIGNAL → CONTEXT → DECISION → ACTION → EFFICACY → INTELLIGENCE
01Technology & AIWhat is running, approved, reliable and worth scaling?
02SecurityWhat can leak, be abused or create a reachable attack path?
03Data & privacyWhich sensitive data is touched, transformed or sent?
04Risk & complianceWhich obligation, exception, control and evidence applies?
05Finance & procurementWhat is adopted, duplicated, valuable or wasteful?
06LeadershipWhere is AI creating value, exposure and accountability?
05 GOVERNANCE CASES / INDIA CONTEXT / SYNTHETIC

Five AI moments. One honest governance record.

A browser prompt, a coding-agent request, a Cowork task, an MCP instruction and a portfolio decision do not expose the same control point. Statefold acts where it is connected, records what it can only observe and leaves unsupported paths visibly opaque.

01 / BROWSER PROMPTCOMPOSER-BLOCKED

A GST reconciliation prompt contains PAN, GSTIN and an internal token.

The supported Gemini composer is one control point. Unusual sends and a missing extension are not silently counted as covered.

  1. 01SURFACEGemini Web · managed Chrome · Pune finance endpointHostname, page, browser process and managed device identify this route.
  2. 02CONTEXTGST reconciliation · PAN/GSTIN values · internal tokenAccount identity is best-effort; device and process evidence stay separate.
  3. 03DECISIONBrowser policy requires a fresh block verdictLocal classifiers return finding categories and severity for this send.
  4. 04ACTIONCancel Enter or the send-button action at the supported composerStatefold blocks this send; it does not claim to rewrite every browser prompt.
  5. 05EVIDENCERule ID + findings + site/path + device-linked eventThe recorded verdict shows why this send was stopped.
Coverage boundary

Enforced at the supported composer. Programmatic sends, broken selectors or an absent extension become observed or opaque.

02 / CODING AGENTPROXY-BLOCKED

A request to fix UPI retries also contains a production credential.

Statefold blocks the secret-bearing network request without terminating the developer’s coding environment.

  1. 01SURFACECoding agent → proxy-honouring OpenAI HTTPS pathThe route and originating process are recorded without claiming a native Codex integration.
  2. 02CONTEXTBengaluru engineer · payments service · production .env credentialDestination and findings join best-effort PID, path, binary hash, signature and Windows user.
  3. 03DECISIONProduction credentials are denied on this external routeThe verdict is scoped to this request and the available process evidence.
  4. 04ACTIONReturn a local block before forwarding; leave the coding environment runningEligible JSON requests may use configured redaction where that route supports it.
  5. 05EVIDENCEHost/path + rule + verdict + prompt preview + process metadataThe record shows which request was stopped and from which process.
Coverage boundary

Enforced only when the client honours the system proxy and TLS inspection works. Pinned or bypassed payloads stay opaque, while connection and process signals may still be observed.

03 / SURFACE IDENTITYMIXED COVERAGE

Claude Chat is a prompt. Claude Cowork is a task with reach.

The same provider does not mean the same surface, permissions or enforceable control point.

  1. 01SURFACEClaude Chat in managed browser ≠ Claude Cowork in Claude DesktopProvider name alone is never treated as the surface.
  2. 02CONTEXTChat: composer text + destination · Cowork: folder, browser and tool requestProcess, file-write and MCP inventory may be visible; the full cloud plan is not assumed.
  3. 03DECISIONAllow the clean Chat prompt · require tighter privilege for CoworkSeparate rules use process, destination, data and available permission evidence.
  4. 04ACTIONStop Chat at the composer when matched · block supported Cowork egressLocal Cowork process, file and MCP activity can remain observation-only.
  5. 05EVIDENCESeparate browser, process, egress, file and MCP recordsEach record preserves what was enforced, observed or unresolved.
Coverage boundary

Remote Cowork execution and connector actions remain opaque unless separately integrated. Chat and Cowork are never presented as one governed surface.

04 / MCP RESPONSERESPONSE-BLOCKED

A provider response instructs an agent to call a tool with a production credential.

On a supported response shape, Statefold stops the sensitive instruction before the agent receives it.

  1. 01SURFACESupported proxied API response → agent → filesystem or PostgreSQL toolProvider endpoint, response direction and tool-call shape identify this route.
  2. 02CONTEXTHyderabad operations agent · sensitive tool argumentsFinding categories join adjacent agent and privilege evidence when available.
  3. 03DECISIONWithhold the response before the client receives the tool instructionThe decision applies only to supported proxied response shapes.
  4. 04ACTIONProxy blocks the response; this instruction cannot initiate the tool callNo claim is made about a separate local or bypassed instruction.
  5. 05EVIDENCEResponse verdict + rule + findings + tool-call markerThe record proves this response was withheld at this control point.
Coverage boundary

Local MCP servers are inventoried by process, configuration and port. Arbitrary local stdio MCP messages remain opaque.

05 / PORTFOLIOHUMAN DECISION

Procurement and the MD decide which AI services to renew, restrict or retire.

Evidence from India operations replaces vendor slides, while the final commercial decision remains human.

  1. 01SURFACEFleet activity + prompt, agent and MCP inventory + gateway tracesOnly tenant-scoped, connected evidence sources contribute.
  2. 02CONTEXTBengaluru engineering · Mumbai finance · Hyderabad operationsOwner, purpose, activity, risk, grant state and named gaps are joined.
  3. 03DECISIONHuman choice: renew, limit or decommissionThe system abstains where activity, ownership or entitlement evidence is missing.
  4. 04ACTIONAdministrator updates model lifecycle, agent privilege or fleet policyStatefold does not claim to buy, cancel or reconcile seats by itself.
  5. 05EVIDENCETimestamped approver + reason + policy version + grant/model historyThe record shows who decided what, when and from which evidence.
Coverage boundary

This is a governed workflow, not complete contract reconciliation. Billing and entitlement completeness require connected source systems.

THE RULE ACROSS EVERY CASE

Enforced means an action and result exist in the event record. Observed means Statefold saw the activity but did not change it. Opaque means the route was not accessible.

02 / Why Statefold compounds

Not six features.
Six reinforcing advantages.

Each layer makes the next one more valuable. Together they form a control and evidence position that grows with every governed interaction.

01
Reach

See beyond one gateway.

Managed endpoints, browser and IDE collectors, agent and MCP inventory, server-side gateway traffic and bounded operational signals.

More governed surfaces
02
Identity

Preserve the actor and surface where evidence identifies them.

Keep person, device, process, agent, tool, privilege, model and route distinct when collector evidence resolves them; unknown or opaque dimensions stay explicit.

Higher-fidelity decisions
03
Context

Join the evidence graph.

Connect AI use to sensitive data, software provenance, vulnerabilities, ownership, business purpose, cost and regulatory obligation.

Compounding enterprise context
04
Action

Turn intent into governed control.

Apply local or gateway decisions where supported: allow, alert, redact, block, route, revoke or require review.

Operational governance
05
Truth

Prove what changed.

Keep intended, configured, deployed, exercised, effective and proven states separate. Coverage gaps remain visible.

No false green
06
Intelligence

Keep the learning sovereign.

Hive and deterministic reasoning explain tenant evidence without silently sending enterprise memory to a hosted model.

Enterprise-owned advantage
03 / The efficacy gate

A control is not real because a screen says “enforce”.

For supported controls with stage-specific receipts, Statefold follows the path from intention to outcome. Leadership gets a truthful answer to the only question that matters: did it work?

  1. 01IntendedThe rule and owner are clear.
  2. 02ConfiguredThe control is set correctly.
  3. 03DeployedIt reached the intended surface.
  4. 04ExercisedA real or tested event invoked it.
  5. 05EffectiveThe expected outcome was confirmed.
  6. 06ProvenThe receipt can withstand review.
Enforced

A supported control applied the decision and confirmed the result.

Observed

Statefold saw the activity but does not claim it changed the request.

Opaque

Unsupported or inaccessible traffic remains an explicit coverage gap.

04 / One AI event, five team views

One AI event. Five teams. Clear next moves.

Leadership, technology, security, risk and finance each see the same event through the decision they own.

MD / Board / Executive Committee

See adoption, value and exposure in the same frame.

Know where AI is creating leverage, where accountability is missing and whether the organisation’s controls are working in practice.

“Where is AI creating value without an accountable owner?”
TOTAL INTELLIGENCELEADERSHIP VIEWILLUSTRATIVE
DECISIONScale what is governed. Intervene where value and accountability diverge.
AdoptionBusiness-aligned
AccountabilityOwner-resolved
Control efficacyEvidence-bound
One evidence graphNo parallel version of truth
05 / Hive Mind

Ask your evidence.
Not the internet.

Hive answers from tenant-owned evidence. It returns a verified fact, an evidence-graph match, or an honest statement that the available evidence is insufficient.

  • No silent hosted-model fallback
  • Citations back to enterprise evidence
  • Deterministic authority remains authoritative
HIVE / TOTAL INTELLIGENCEIllustrative
Demo boundary

This deterministic walkthrough shows how Hive separates evidence-backed answer types from insufficient evidence. It does not read tenant data or issue live citations.

Illustrative deterministic interaction · no network request

06 / Architecture

Sovereign by design.
Open by necessity.

Statefold sits across the enterprise AI estate without forcing the enterprise into one model, cloud or security stack.

01 / SIGNAL

Enterprise AI estate

Managed endpoints, browsers, IDEs, agents, MCP, gateways, models, files and bounded system signals.

02 / CONTROL

Statefold

Identity, inventory, policy, risk, decisions, actions, efficacy and tenant boundaries.

03 / INTELLIGENCE

Evidence + Hive

Canonical evidence, relationship projections, grounded answers and verifiable outcomes.

Self-hosted control

Enterprise evidence and Hive reasoning stay inside the chosen boundary.

Provider-neutral gateway

Approved application requests route only to explicitly configured providers.

Tenant isolation

Transactional state and evidence remain tenant-scoped by design.

Extensible surface

APIs and evidence exports connect identity, SIEM, GRC, data and cloud ecosystems.

Built to connect with
Cloud + modelsIdentitySIEM + SOARGRCEndpointDataAPIs
The platform opportunity

Govern the system.
Prove the outcome.

Bring technology, security, risk, compliance and leadership into one accountable operating picture.

Request an executive briefing ↗ Review the architecture ↗