Connect prompts & outputs
Illustrate how to wire guardrails before a prompt node or after an LLM output, including optional projects.
TestSavant Guardrails for n8n
Ship safe automations without writing security glue
Drag TestSavant guardrail nodes onto your canvas, capture API credentials securely, pick scanners, reuse project defaults, and connect to any LLM prompt or output in minutes.
Secure credentials
Curate risk coverage
Reusable configs with projects
Prompt & output hooks
Integration snapshots
Overlay guardrails on every n8n surface
Replace each placeholder with your own screenshots to show operators exactly what to configure.
Capture TestSavant API keys
Display the credentials drawer with scoped key creation, rotation reminders, and audit visibility.
Pick scanners per flow
Highlight toggles for injection, leakage, safety, and compliance coverage with risk presets.
Review evidence & analytics
Show the audit trail panel that captures scanner results, reasons, and remediation guidance.
Setup essentials
Four fields to wire once and reuse everywhere
1. API key credentials
Paste TestSavant API keys into the secure credential helper. Each key can be limited to read, run, or configure guardrails per project.
- • Scoped keys per environment (dev, staging, prod)
- • Automatic rotation reminders and usage analytics
- • Vault storage or customer-managed secret stores
2. Choose scanners
Select which TestSavant scanners run on every request—prompt injection, toxicity, policy, data leak, and custom checks.
- • Preset bundles for support bots, ops automations, or RAG flows
- • Override weighting per guardrail node
- • Inline tooltips describing each scanner
3. Optional default project
Tie each node to a TestSavant project so policies, scanners, and evidence routing stay consistent whenever you duplicate or update flows.
- • Projects hold scanner configs, policies, and model inventory
- • Workspace-level defaults for fast duplication
- • Project tags for routing evidence to the right team
4. Prompt & output connectors
Wire guardrails directly to your LLM prompt input or response output. Support for text, tools, and structured payloads keeps every channel safe.
- • Works with OpenAI, Anthropic, Azure, OSS LLMs, and custom agents
- • Optional output transforms for redaction or citation injection
- • Stream responses after guardrails approve or sanitize
Why in-flow guardrails
Stop incidents inside your automation
Guardrails sit between triggers, LLM calls, and downstream actions. Risky prompts are neutralized before they reach the model; unsafe outputs are transformed before anyone sees them.
- Block prompt injection & jailbreaks: Inspect every prompt for intent shifts, tool pivots, and leaked policy instructions.
- Redact secrets automatically: Scrub PII, PHI, access tokens, and customer data before they leave your systems.
- Gate high-impact actions: Require approvals when automations want to email, sync, or post on behalf of humans.
Built for n8n speed
Drop nodes, publish flows, ship faster
Swap manual scripts for TestSavant guardrail nodes or webhooks. The platform handles scanners, policies, telemetry, and evidence without slowing you down.
Guardrail: Input
Scan prompts before your LLM, agent, or tool node executes.
Guardrail: Output
Transform or block responses prior to human delivery or tool chaining.
Decision webhook
Keep flows async-friendly with lightweight allow / transform / block calls.
Evidence packs
Export clause-mapped proof for auditors, customers, or internal review.
Operational flow
How guardrails wrap an n8n automation
Use this diagram along with your own canvas screenshots to explain the flow from trigger to evidence.
High-impact use cases
No-code automations that stay secure
Customer support copilot
Sanitize queries, enforce disclaimers, and block account‑takeover or data‑exfil attempts in chat flows.
Ops automations
Gate risky actions (send‑email, write‑file, post‑message) via webhooks; require approvals automatically.
RAG pipelines
Tenant‑scoped retrieval; minimum‑necessary access; require citations or transform responses.
Frequently asked questions
Common queries about guardrails
Will guardrails slow my flows?
They run in a low‑latency path designed for online traffic. Deep checks can be scoped per route.
Do I need to code?
No. Use our n8n custom nodes and simple webhooks—drag, drop, and go.
Is it model‑agnostic?
Yes. It works with major LLM providers, OSS runtimes, and tools invoked inside n8n.
What about compliance?
Decisions and lineage are logged with evidence IDs; export clause‑mapped packs for audits.
Can I deploy privately?
Yes. VPC deployment and customer‑managed keys are supported; evidence can mirror into your trust portal.
Secure your no‑code automations today
Inline guardrails, instant observability, adaptive defense. Build faster—and prove it’s safe.
