TrustFabric help center
How TrustFabric works, how to connect agents and pipelines, and what to do when a decision is not what you expected.
Documentation version 1.0 · Applies to TrustFabric · Updated 2026-10-30
Articles marked Planned describe capabilities that are not built yet. About TrustFabric
36 articles
Getting Started
- What is Clearception TrustFabric?Available
Enterprise AI engineering governance: your existing policies, applied to every governed agent and pipeline action, with evidence.
- How TrustFabric worksAvailable
Policy documents become draft rules, people approve a bundle, the gateway evaluates requests against it, and each decision becomes evidence.
- Subscription and plan comparisonAvailable
Starter, Growth, Business and Enterprise differ by seats, repositories, monthly evaluations and capabilities.
- Organization setupAvailable
Create the TrustFabric organization, assign the people who author and approve policy, and define your first jurisdiction.
- Connecting your first repositoryAvailable
Register a repository with its classification and jurisdiction, then add a pipeline gate step.
Developer Integrations
- Claude Code integrationAvailable
A PreToolUse hook script asks the gateway before each tool call. Enforcement level: voluntary.
- Claude Agent SDK integrationAvailable
Route every tool call through a can_use_tool callback that calls the gateway inside your own runner. Enforcement level: enforced.
- Codex governance integrationAvailable
Inventory OpenAI Codex agents and accept their gateway calls. Monitored or voluntary only — no mediated path.
- GitHub integrationAvailable
A GitHub Actions step calls the pipeline gate with a workload token; anything other than ALLOW fails the job.
- Azure DevOps integrationAvailable
An Azure DevOps script task calls the same pipeline gate. The Azure DevOps wiki policy connector is planned.
- MCP connector configurationPlanned — not built yet
A packaged TrustFabric MCP connector is not built yet. MCP-based agents can call the HTTP gateway directly.
Security and Governance
- Policy authorization architectureAvailable
One deterministic gateway, one approved bundle version per decision, and fail-closed defaults.
- Agent permissionsAvailable
The agent inventory: identity, tool permissions, approved repositories and environments, risk class, kill switch and an honest enforcement level.
- Tool execution controlsAvailable
Restrict which tools an agent may use, on which repositories and environments, and with which model endpoints.
- Enterprise authenticationAvailable
People sign in through Clearception platform SSO with your enterprise identity provider (Business plan and up). Agents and CI use workload tokens.
- Tenant isolationAvailable
Starter, Growth and Business run on shared multi-tenant infrastructure, isolated at the organization boundary.
- Secret managementAvailable
Workload tokens are shown once and stored hashed. Store them in your CI secret store.
Compliance and Sovereignty
- Data sovereignty configurationAvailable
Define jurisdictions with allowed regions, approved model endpoints, transfer rules by classification and retention.
- Regional processing rulesAvailable
Requests that would process data outside allowed regions are denied. No regional execution environments are provisioned yet.
- Policy ingestion and approvalAvailable
Upload or paste policy text with provenance; review draft rules; approve bundles under the two-person rule.
- Legal controlsAvailable
Map contractual and regulatory obligations to rules, with approvals and time-boxed exceptions decided by people.
- Evidence generationAvailable
Every decision becomes a tamper-evident record in a per-organization HMAC hash chain, exportable as JSON.
- Compliance reportingAvailable
Export a JSON audit package today. Framework-formatted reports and audit-system integrations are not built.
Billing and Administration
- Subscription managementAvailable
Buy Starter, Growth or Business from the console; manage payment details, invoices and cancellation in the billing portal.
- Developer seatsAvailable
Each plan includes a number of governed developers; more can be added per seat.
- Repository allowancesAvailable
Each plan includes a number of governed repositories; more are added in blocks of 100.
- Policy evaluation usageAvailable
One evaluation is one auditable policy decision. Usage is metered monthly with notifications at 80% and 100%.
- Add-ons and overagesAvailable
Overage is billed only after you agree to it. Without agreement: a 10% grace allowance, then new registrations pause while authorization keeps running.
- Enterprise contractsAvailable
Enterprise capacity, infrastructure, SLAs and onboarding are set by contract.
Troubleshooting
- Agent action deniedAvailable
Read the reasons on the decision, then check the bundle, the agent's permissions and the kill switch.
- Missing policy approvalAvailable
Draft rules do nothing until a bundle containing them is approved by someone other than its author.
- Unsupported regionAvailable
Operations that need a regional execution environment are denied because none is provisioned yet.
- Failed repository connectionAvailable
Check the repository registration, the token and the plan's repository allowance.
- Pipeline policy failureAvailable
A non-ALLOW gate answer fails the job. The response lists which required checks were missing.
- Subscription entitlement errorAvailable
A feature outside your plan, or a paused registration, returns an entitlement error.
- Missing audit evidenceAvailable
Only actions that reached the gateway have evidence. Check enforcement level and chain verification.