START HERE

Welcome to Teleon

The trust layer for AI agents. Three integration paths. One audit trail. One badge.

Teleon adds policy controls and evidence collection around the AI agents you already operate. You keep your framework and provider, then connect through the Gateway, a framework SDK, or an SDK with the Sidecar. The evidence Teleon can collect and the actions it can enforce depend on the integration path you choose.

How Teleon fits around an agent

1
Your agent receives a requestYour existing application, framework and model continue to run the workflow.
2
The integration reports evidenceTeleon receives only the model, framework, tool, entity or custom-signal evidence visible at the selected boundary.
3
An available evaluator applies policyGateway and Sidecar can return supported allow, redact, hold or block decisions. Middleware records activity but does not evaluate policy.
4
The outcome becomes evidenceEvents, policy decisions and available request context enter the agent's audit trail and contribute to later trust evidence.

Example: governing an application-computed payment risk. A payment agent calculates a velocity score and publishes it with trust.emit(). The SDK records the signal. When that SDK is connected to Sidecar, a policy can evaluate the score before a later send_payment call and block the action while preserving the signal and decision in the audit ledger. Gateway cannot see this application-owned signal because it exists outside the model boundary.

Choose an integration boundary

01GatewayChoose this when supported provider traffic can be routed through Teleon and model-boundary policy decisions are sufficient.Explore Gateway →02Framework SDKChoose this when you need in-process evidence for actual tool execution, agent flow and framework events without local policy enforcement.Explore SDKs →03SDK + SidecarChoose this when policies must evaluate framework evidence, entities or custom signals before supported application actions run.Explore Sidecar →

What Teleon provides

DISCOVERYTrust ProfileVersioned evidence observed at the integrated boundary.

INTEGRITYAudit ledgerSigned and hash-chained, with external anchoring when configured.

CONTROLPolicy compilerNatural-language rules validated against supported capabilities.

MEASUREMENTTrust ScoreA 0–100 result calculated from the evidence available for the agent.

TESTINGRed-team evidenceScheduled testing for eligible agents with sufficient shadow-twin evidence.

ASSURANCEBundles and BadgeSigned evidence packages and a revocable badge for eligible agents.

Your path through the documentation

1
Take the tutorialLearn the dashboard with ten clearly labelled synthetic calls.
2
Connect your agentChoose a supported integration path and send real evidence.
3
Review what Teleon observesConfirm the profile, audit events and integration boundaries.
4
Configure policiesDeploy only rules that the selected topology can evaluate.

NEW WORKSPACEStart with a safe, guided walkthrough

The tutorial uses synthetic events and does not connect to your production agent.

Open the first-run tutorial →

Scope and limitations

How this documentation is organized

The left sidebar follows the recommended learning path. Start with the First-run tutorial, use Quickstart to connect a real agent, and read Core concepts for the shared vocabulary. Continue to Integration to select a deployment path, Trust to understand profiles, policies and scoring, and Audit & Compliance for ledger, DSR, bundle and badge behavior. Reference contains the HTTP API, CLI and changelog.