Spctre covers the full agent governance lifecycle — set policy rules, test changes safely against history, enforce decisions at runtime or observe first, and maintain audit-ready records — without requiring you to change your framework, cloud, or model.
Every policy change lives on a branch with authorship, diff, timestamp, rationale, and target workspace. Teams can author policy files in Git repositories and import drafts via CI/CD, or manage policy branches directly in Spctre.
Every policy change lives on a branch with authorship, diff, timestamp, rationale, and target workspace. Teams can author policy files in Git repositories and import drafts via CI/CD, or manage policy branches directly in Spctre.
Require approval before a policy goes live. Approval state, reviewer identity, rationale, escalation status, and SLA timing are all recorded as durable audit entries.
Require approval before a policy goes live. Approval state, reviewer identity, rationale, escalation status, and SLA timing are all recorded as durable audit entries.
Org-level, workspace, environment, and connector policies compose into one versioned, hashed bundle targeted to specific runtimes — so you can prove exactly which policy was active at the time of any decision.
Org-level, workspace, environment, and connector policies compose into one versioned, hashed bundle targeted to specific runtimes — so you can prove exactly which policy was active at the time of any decision.
Import connector packs, framework packs, enterprise baselines, and local overlays. Source, version, and compatibility metadata are preserved so you know where every rule came from.
Import connector packs, framework packs, enterprise baselines, and local overlays. Source, version, and compatibility metadata are preserved so you know where every rule came from.
Compare proposed rules against the active bundle before publishing. Spctre surfaces conflicts and identifies the exact pack, branch, rule, and reviewer path involved.
Compare proposed rules against the active bundle before publishing. Spctre surfaces conflicts and identifies the exact pack, branch, rule, and reviewer path involved.
Block publish when required reviews, simulations, branch checks, or compliance conditions are missing — and produce an audit log for each blocked attempt so there's a record of what was missing and when.
Block publish when required reviews, simulations, branch checks, or compliance conditions are missing — and produce an audit log for each blocked attempt so there's a record of what was missing and when.
A policy governs an action. A Blueprint governs an agent. Each one states the tasks, tools, connectors, budgets, approvals, and runtime targets a single agent is permitted — reviewed and published before that agent is allowed to act.
Permitted tasks, tools, connectors, budgets, approval requirements, and runtime targets are bound to a specific policy revision in a single document — so a reviewer can see an agent's whole intended scope in one place instead of inferring it from scattered rules.
Permitted tasks, tools, connectors, budgets, approval requirements, and runtime targets are bound to a specific policy revision in a single document — so a reviewer can see an agent's whole intended scope in one place instead of inferring it from scattered rules.
Blueprints move through draft, review, and publish with recorded approvals and rollback. Approvals require an explicit reviewer decision while the Blueprint is in review — the state and the reviewer are both part of the durable record.
Blueprints move through draft, review, and publish with recorded approvals and rollback. Approvals require an explicit reviewer decision while the Blueprint is in review — the state and the reviewer are both part of the durable record.
Published Blueprints compile to a runtime contract. Gateway decisions bind the matching agent and policy revision, then stop connector or tool actions the Blueprint never declared — including ones a model improvises at runtime.
Published Blueprints compile to a runtime contract. Gateway decisions bind the matching agent and policy revision, then stop connector or tool actions the Blueprint never declared — including ones a model improvises at runtime.
Keep Blueprint sources in Git and import them from CI with an operator key. Validate locally with a dry run, then import as a draft — the import never approves or publishes on its own.
Keep Blueprint sources in Git and import them from CI with an operator key. Validate locally with a dry run, then import as a draft — the import never approves or publishes on its own.
Capture ALLOW, DENY, and REVIEW records with connector, action, agent identity, policy references, artifact hash, environment, and runtime target through supported adapters, gateway decisions, or mapped evidence integrations.
Capture ALLOW, DENY, and REVIEW records with connector, action, agent identity, policy references, artifact hash, environment, and runtime target through supported adapters, gateway decisions, or mapped evidence integrations.
Gateway decisions, tool calls, human approvals, policy bundle hashes, and runtime targets are linked into a single inspectable ledger — not scattered across separate log sources.
Gateway decisions, tool calls, human approvals, policy bundle hashes, and runtime targets are linked into a single inspectable ledger — not scattered across separate log sources.
Audit ledgers are maintained with hash-chain integrity. Verification checks detect unexpected changes to historical records, so you can trust what the log says.
Audit ledgers are maintained with hash-chain integrity. Verification checks detect unexpected changes to historical records, so you can trust what the log says.
Configure retention windows by workspace and connector. Redact sensitive values before the record is written — preserving the structural log and provenance without retaining the sensitive content.
Configure retention windows by workspace and connector. Redact sensitive values before the record is written — preserving the structural log and provenance without retaining the sensitive content.
Confirm ledger continuity, hash-chain integrity, expected policy references, and audit shape before export or incident review — so you're not discovering problems during an audit.
Confirm ledger continuity, hash-chain integrity, expected policy references, and audit shape before export or incident review — so you're not discovering problems during an audit.
Each decision retains the exact published policy bytes it was evaluated against and the version of the evaluator that produced it. One deterministic engine decides in the control plane, the runtime service, and portable hosts — so a simulation and a production decision on the same inputs don't drift apart, and an old decision can still be reproduced after an upgrade.
Each decision retains the exact published policy bytes it was evaluated against and the version of the evaluator that produced it. One deterministic engine decides in the control plane, the runtime service, and portable hosts — so a simulation and a production decision on the same inputs don't drift apart, and an old decision can still be reproduced after an upgrade.
Query by agent, connector, action, environment, decision status, policy reference, reviewer, workspace, runtime target, or artifact hash. Quickly locate decisions across large operational history.
Query by agent, connector, action, environment, decision status, policy reference, reviewer, workspace, runtime target, or artifact hash. Quickly locate decisions across large operational history.
Stream policy evidence events to Splunk or Microsoft Sentinel in real time, so security teams can correlate governed agent activity with their existing detection and response workflows.
Stream policy evidence events to Splunk or Microsoft Sentinel in real time, so security teams can correlate governed agent activity with their existing detection and response workflows.
Run proposed policy changes against past production events and see exactly which decisions would flip — before the policy is published and before any agent is affected.
Run proposed policy changes against past production events and see exactly which decisions would flip — before the policy is published and before any agent is affected.
See which workflows a new policy would block, approve, or escalate, broken down by connector, workspace, environment, and runtime target.
See which workflows a new policy would block, approve, or escalate, broken down by connector, workspace, environment, and runtime target.
Generate structured packets for security reviews, customer incident responses, and regulatory inquiries — including policy references, artifact hashes, reviewer chain, timestamps, and decision records.
Generate structured packets for security reviews, customer incident responses, and regulatory inquiries — including policy references, artifact hashes, reviewer chain, timestamps, and decision records.
Map decision evidence and control logs to support SOC 2, ISO 27001, HIPAA, GDPR, PCI DSS, NIST AI RMF, and public-sector audit preparation. The same evidence records map to multiple framework controls without duplicating governance work.
Map decision evidence and control logs to support SOC 2, ISO 27001, HIPAA, GDPR, PCI DSS, NIST AI RMF, and public-sector audit preparation. The same evidence records map to multiple framework controls without duplicating governance work.
Compare active policies against expected baselines and surface changed controls, missing approvals, stale bundles, and connector-specific drift before they become a compliance finding.
Compare active policies against expected baselines and surface changed controls, missing approvals, stale bundles, and connector-specific drift before they become a compliance finding.
Reviewer chain, decision rationale, timestamps, branch revisions, simulation outcome, and publish gate state are all preserved in a format an auditor can inspect without direct system access.
Reviewer chain, decision rationale, timestamps, branch revisions, simulation outcome, and publish gate state are all preserved in a format an auditor can inspect without direct system access.
Evaluate agent actions against your active policy bundle in real time. Returns ALLOW, DENY, or REVIEW with policy references, trace IDs, and latency telemetry — designed for low-overhead inline evaluation.
Evaluate agent actions against your active policy bundle in real time. Returns ALLOW, DENY, or REVIEW with policy references, trace IDs, and latency telemetry — designed for low-overhead inline evaluation.
REVIEW decisions pause the workflow and route the action to a named approver. The approver sees full decision context, approves or rejects, and the outcome — including their identity and rationale — is recorded in the ledger.
REVIEW decisions pause the workflow and route the action to a named approver. The approver sees full decision context, approves or rejects, and the outcome — including their identity and rationale — is recorded in the ledger.
Track trust scores and context budget consumption per agent across sessions. Violations surface as governance events with full ledger provenance, so you know which agent crossed which threshold and when.
Track trust scores and context budget consumption per agent across sessions. Violations surface as governance events with full ledger provenance, so you know which agent crossed which threshold and when.
Record governance decisions without halting runtime behavior. Use observed audit logs to tune policies and build confidence before switching to enforcement.
Record governance decisions without halting runtime behavior. Use observed audit logs to tune policies and build confidence before switching to enforcement.
Block denied actions, route review actions to approvers, and return decision metadata to the calling runtime. Move individual agents from observe to enforce independently, at your own pace.
Block denied actions, route review actions to approvers, and return decision metadata to the calling runtime. Move individual agents from observe to enforce independently, at your own pace.
Support device authorization, refresh token rotation, service account keys, and scoped credentials — so agents can run continuously without requiring overly broad permissions.
Support device authorization, refresh token rotation, service account keys, and scoped credentials — so agents can run continuously without requiring overly broad permissions.
Policies, audit logs, identities, service accounts, and workspaces are all scoped to tenant boundaries. Tenant-scoped access controls strictly isolate data at the query and storage layers.
Policies, audit logs, identities, service accounts, and workspaces are all scoped to tenant boundaries. Tenant-scoped access controls strictly isolate data at the query and storage layers.
Connect your enterprise identity provider (SAML 2.0 and SCIM automated provisioning available on Enterprise plans; OIDC on Cloud), inherit your identity provider’s SSO and MFA policies, configure role-based access, and manage members and administrative review workflows.
Connect your enterprise identity provider (SAML 2.0 and SCIM automated provisioning available on Enterprise plans; OIDC on Cloud), inherit your identity provider’s SSO and MFA policies, configure role-based access, and manage members and administrative review workflows.
Spctre captures decision context — agent, connector, action, policy match, outcome — without storing action payload data unless you explicitly configure it. Sensitive values can be redacted at the gateway before any record is written.
Spctre captures decision context — agent, connector, action, policy match, outcome — without storing action payload data unless you explicitly configure it. Sensitive values can be redacted at the gateway before any record is written.
Manage API tokens, service keys, refresh flows, revocation events, and scoped access for automation and agent runtimes — with audit records for every credential lifecycle event.
Manage API tokens, service keys, refresh flows, revocation events, and scoped access for automation and agent runtimes — with audit records for every credential lifecycle event.
Route review queues, policy drift, suspicious decisions, gateway failures, and compliance events to the channels your team already monitors — Slack, Microsoft Teams, PagerDuty, or webhook.
Route review queues, policy drift, suspicious decisions, gateway failures, and compliance events to the channels your team already monitors — Slack, Microsoft Teams, PagerDuty, or webhook.
Run Spctre in managed cloud, deploy privately into your own infrastructure, or use local and custom adapter patterns for strict data residency or air-gapped requirements.
Run Spctre in managed cloud, deploy privately into your own infrastructure, or use local and custom adapter patterns for strict data residency or air-gapped requirements.