Documentation

Keeptrusts Docs

Keeptrusts helps teams route, govern, and review AI traffic without rebuilding every application. Define runtime behavior in config, run traffic through the gateway, and use the product surfaces to verify what changed.

Get startedHow it works

Quickstart

Install the CLI, create a starter config, and send your first governed request.

Config-first workflow

Use policy-config.yaml as the source of truth for policy, routing, and evidence behavior.

Policy controls

Choose the controls you need for prompt safety, PII handling, quality checks, routing, and human review.

Install the gateway

Run Keeptrusts locally or in your own environment and point applications at the governed endpoint.

---

A reliable first session

Use the docs in the same order that traffic moves through the product:

1. Install the CLI and gateway and confirm the process is healthy.
2. Create and validate a policy config before starting the gateway with that file.
3. Send an authenticated request through a supported request family.
4. Confirm both the client response and the applicable runtime evidence. Use kt events tail for delivered request outcomes, History only for capture-enabled session context, Trail for audit chronology, and Inbox for human-review work.

/healthz proves that the gateway process is running. It does not prove that a client token is authorized, a model is available, or a provider credential is valid. Use an authenticated model request for that end-to-end check.

Keep the credentials separate

Credential	Used by	Purpose
Gateway runtime API token	kt gateway run --agent ...	Connects the gateway runtime to the Keeptrusts control plane
Client API token	Application, SDK, or IDE	Authorizes a request sent to the gateway
Upstream provider credential	Gateway provider target	Authorizes the gateway with the selected model provider

Do not reuse the gateway runtime token as the client or provider credential. The setup and rotation workflows are different for each boundary.

Documentation

Use the rest of the site by job:

Configuration reference

Field-by-field reference for declarative config, providers, routes, tests, and policy settings.

CLI reference

Run the gateway, validate config, inspect runtime signals, and automate rollout workflows.

Integrations

Connect OpenAI-compatible clients, provider SDKs, IDEs, and existing application stacks.

Operational workflows

Follow the customer tasks for onboarding, rollout, investigation, evidence export, and spend review.

Next steps

• Quickstart
• Config-First Workflow
• Customer Journeys