SIDERUNE

SIDERUNE Product

Accountability infrastructure for AI agents that take real-world action.

SIDERUNE connects delegated authority to agent actions, actual effects, and verifiable evidence — so you can reconstruct what happened and determine whether the resulting effect stayed within the authority that was granted.

Explore the Example Request a Technical Demo

Product role

Authorization decides what should be allowed. Accountability proves what actually happened.

Existing systems may authenticate identities, define permissions, evaluate policies, expose tools, or execute payments.

SIDERUNE sits beside those systems. It reconstructs the chain between recorded authority and actual effect, then preserves the evidence needed to verify that reconstruction.

  1. Identity
  2. Authority
  3. Delegation
  4. Action
  5. Effect
  6. Evidence

Actor identities are recorded as supplied; SIDERUNE does not authenticate agents.

The result is an accountability record connecting the recorded authority chain to what ultimately happened in the real world.

How SIDERUNE works

From events to verifiable accountability.

  1. Record the authority context

    Record the identity, authority and delegation events supplied to SIDERUNE that establish the authority context for an agent.

  2. Connect the action

    Link the agent's requested action to the recorded authority and delegation chain that existed when the action occurred.

  3. Record the observed effect

    Ingest the effect reported from the external system so the requested action can be distinguished from the recorded real-world result.

  4. Reconstruct and compare

    Rebuild the causal chain and compare the recorded effect with the authority that applied to the agent.

  5. Produce verifiable evidence

    Preserve the resulting evidence in a tamper-evident, cryptographically verifiable chain.

A reconstruction is only as complete as the events supplied to SIDERUNE.

If no effect event is supplied, SIDERUNE can only conclude that no effect was recorded in the supplied evidence — not that no real-world effect occurred.

Simulated accountability scenario

A $100 authority. A $482 effect.

A human grants Agent A authority to issue refunds up to $100. Agent A delegates authority to Agent B. Agent B requests a $482 refund. The external system records an actual $482 effect.

Delegated authority
$100
Actual effect
$482
Authority exceeded
$382
Evidence
VERIFIED
  1. Human grants authority
  2. Authority ≤ $100
  3. Agent A holds authority
  4. Delegated to Agent B inherits ≤ $100
  5. Requested $482
  6. Actual effect $482
  7. EXCEEDED violation
  8. Evidence Verified hash chain intact

Authorized limit $100. Requested $482. Actual effect $482, $382 beyond the limit. Violation: true. Root cause: the delegated agent exceeded its inherited authority — reconstructed and verified against the evidence chain.

SIDERUNE reconstructs the recorded authority chain, connects it to the action and observed effect, identifies the authority violation, and preserves the evidence required to verify the reconstruction.

Integration boundary

Designed to sit beside your existing control plane.

Your existing execution path

  1. IAM / RBAC / Policy / MCP
  2. AI Agent
  3. Application / Payment / External System

Events supplied to SIDERUNE

SIDERUNE accountability layer

  1. Authority reconstruction
  2. Effect comparison
  3. Evidence
  4. Accountability
Conceptual diagram. SIDERUNE is not in the execution path and does not intercept requests: it receives events supplied from the existing stack and reconstructs accountability beside it.

SIDERUNE does not replace:

IAM · RBAC · policy engines · MCP servers · payment infrastructure · application-level authorization

You keep your existing execution and authorization architecture. SIDERUNE adds an independent accountability layer beside consequential agent actions by reconstructing the events supplied to it.

What SIDERUNE produces

Evidence you can inspect and verify.

  • Authority lineage

    Who authorized whom and how authority was delegated in the supplied event chain.

  • Action reconstruction

    What action was requested and the recorded authority context attached to it.

  • Actual-effect comparison

    What the supplied external-system effect reports happened versus what the agent was authorized to do.

  • Authority verdicts

    Whether the recorded effect remained within the reconstructed authority.

  • Tamper-evident evidence

    A cryptographically verifiable evidence chain for the reconstructed sequence.

  • Forensic reconstruction

    A causal record that can be inspected after consequential agent activity.

Current product boundary

Clear boundaries by design.

SIDERUNE currently provides

  • Authority and delegation reconstruction
  • Action-to-effect correlation
  • Authority-limit comparison
  • Accountability verdicts
  • Tamper-evident evidence
  • Cryptographic verification

SIDERUNE does not currently claim to provide

  • A managed runtime policy-enforcement gateway
  • A replacement for IAM or authorization infrastructure
  • A payment execution layer
  • Complete reconstruction when required events were never supplied

SIDERUNE's cryptographic and engineering properties do not by themselves imply regulatory certification or compliance.

Technical walkthrough

When an AI agent takes consequential action, can you reconstruct and prove whether it was authorized?

See how SIDERUNE connects authority, delegation, action, effect, and evidence in a technical walkthrough.

Request a Technical Demo