For agent and platform engineers

Govern actions without judging every call.

Check proposed tool calls before execution. Resolve explicit policy conditions in code, and request model evidence or human review only when the policy requires it.

5D is an action-policy gate for AI agents: it evaluates declared tool actions before the host executes them.

Explore the action flow

Current assurance boundary

5D provides a versioned deterministic runtime-control layer. When an integration supplies the required action identity and outcome data, release-specific decision records can support reconstruction of declared governed actions.

Use a versioned action contract, explicit policy results, and replayable evidence to debug and verify the path before a side effect.

The records show the properties tested, the assumptions made, and the limitations that remain. The integrating organisation retains responsibility for identity, execution, outcome capture, path coverage, and any legal or assurance conclusion.

Workload and collaboration

Keep clear actions on an inspectable policy path.

Use versioned policy for declared action envelopes, then route the cases your policy selects to optional Jev, another observer, or human review.

Proposed action Typed action envelope

Tool, target, purpose, principal, and context enter one declared contract.

Deterministic policy Evaluate before side effect

Versioned rules resolve declared inputs on documented integration paths.

Policy-routed evidence Optional Jev or observer

Observer evidence can tighten or hold. It cannot clear an explicit block.

Illustrative workload model

Change the declared assumptions.

Declared assumptions, not measured 5D performance.

Total proposed actions per day Actions resolved by policy alone (%) Currency Observer cost per routed request

365-day illustrative year. Currency is a display/input unit; no exchange-rate conversion is applied.

Deterministic-only actions / day6,500

Observer- or review-routed actions / day3,500

Estimated routed observer spend / year€6,387.50

One collaboration layer

Five owners, one versioned action contract.

Policy intent, implementation, threat review, human disposition, and reconstructable evidence stay connected without transferring organisational accountability to a model.

Start from public primitives: inspectable policy evaluation, verdicts, documented adapters and gateways, and evidence records. The support matrix states what version 0.7.0 verifies and what the host still owns.

Inspect setup ↗Read the support matrix ↗

Routing share depends on your policy and traffic.

Observer cost is user-supplied; it is not a 5D price or forecast.

Estimated variable observer spend excludes deterministic infrastructure, human review, retries, and other deployment costs.

Model cost and latency vary with provider, model, caching, concurrency, and deployment design.

Relevant deployment example

Follow one action relevant to your role.

Synthetic tested action under the pinned 5D 0.7.0 illustrative policy.

Cloud operations agent

Apply an infrastructure plan assembled from retrieved instructions

Untrusted context steers the plan toward deleting live resources and snapshots. Under the pinned illustrative policy, the expected result is that the action is ineligible for execution.

D2T4R4E4A3

Do not execute · Score 2.55

Tool: terraform.apply
Target: cloud account
Purpose: repair infrastructure

5D suppliesA deterministic RED decision and rationale.

Host ownsPlan validation, complete interception, and execution refusal.

Coverage limit5D does not inspect every Terraform semantic or prove recovery.

Your evaluation lens

Inspect the action envelope, execution path, latency assumptions, failure behavior, and replayable receipt.

Integration contract

Typed action boundary

Normalize the tool, target, purpose, data class, identity, and delegation context before evaluation.

Latency and compute

Deterministic fast path

Keep explicit policy decisions out of the model path and reserve model calls for genuine ambiguity.

Operational clarity

Explicit failure behavior

Make invalid inputs, engine errors, review branches, and blocked actions visible at the integration boundary.

Debug and test

Replayable receipts

Reconstruct the inputs, policy version, evidence, decision, and simulated execution status.

Evidence your role can use

See the records behind the runtime path.

Test behavior, bind a decision to policy content, and reconstruct one evaluated action.

Test output

Test report

It makes the tested behavior reviewable before policy is promoted or changed.

Runtime record

Decision record

It lets teams reconstruct which policy path evaluated one proposed action and what 5D decided.

Policy provenance

Policy digest

It distinguishes the policy content actually used, even when a human-readable version label is reused.

See every artifact across the lifecycle

Relevant use cases

Start with the action paths you own.

View all use cases ↗

Use case 01 · privileged operations