Inspect untrusted content before it reaches your agent. Run Trismag locally, use the hosted API, or connect it to your workflow.
From untrusted input to an informed decision.
Live use is subject to the terms and privacy policy. Avoid sensitive content.
Edit the text, then inspect it. Sends this input to Trismag for a limited risk preview; we check it in memory and don't keep it. Avoid secrets and personal data.
Instruction override and system-prompt extraction. Hold this input for review.
Example policy shown. Your application decides.
Your key stays in this tab. A live check uses your allowance and sends your text through our gateway to Trismag.
Inspect untrusted content before it reaches your agent. Choose where Trismag runs, then decide how your workflow uses the result.
Keep scan content on your own machine. Use Trismag from Python or your terminal, with no API key or network call during inspection. Download the free local demo now; everyone gets the same updates through October.
Download and set up locallySend content from any language over HTTPS. Start with the live demo or a free key; hosted usage limits apply.
Set up the APIRequire inspection in an n8n workflow, or make scanning available through MCP. Choose who controls the check.
Explore the integrationsLocal and hosted releases share the detector source. Check the engine version when comparing results. A connection alone does not enforce inspection.
Your agent was asked to summarize a page. Somewhere in that page, someone asked it to do something else.
Prompt injection lives inside the content agents need to read. Trismag checks for known signs of instruction overrides, credential requests, hidden characters, and deceptive tool descriptions.
Inspect the evidenceInstruction override detected.
Review before the agent continues.
Make inspection a step in your pipeline.
Give your agent useful information.
Keep the final policy in your hands.
Pages, messages, documents,
and tool metadata.
Trismag measures signals.
Your workflow applies policy.
Allow, review, or block.
With permissions still enforced.
A LOW verdict means no known pattern matched. It is never a permission slip.
Inspect retrieved content for known attacks, encoded text, invisible characters, and authority claims.
Check messages together to surface patterns that emerge across the recent conversation window.
Inspect descriptions and nested schema text for known manipulation. Tool capability and tool trust are different questions.
A useful security tool should explain its findings—and be just as clear about what it misses.
Explore the trust centerFixed rules, no LLM interpretation. The same input and engine version produce the same result.
Local inspection keeps content on your machine. Hosted inspection checks it in memory: we don't keep the text or its verdict. Choose the data path that fits your workflow.
Novel phrasing can pass. We distinguish regression tests from real-world accuracy and publish the limitations.
Python or CLI on your machine. HTTPS from any language. An MCP connection makes scans available; your workflow must enforce them. Understand the difference
# Inspect before your agent reads
curl https://api.trismag.dev/v1/triage \
-H 'Content-Type: application/json' \
-d '{"text":"Ignore all previous instructions"}'
# Abbreviated response · preview tier
{
"risk": "CRITICAL",
"tier": "preview"
}Try Trismag in a real workflow.
Help shape what comes next.
Free demo through November 30, 2026. Everyone uses the same current detector release during the demo. Download Python and CLI inspection for unlimited offline scans, or try the hosted API under its published limits. No card and no automatic enrollment. Download and install
$0No card required
Unlimited local scans. After the demo, Community is planned to receive quarterly detection snapshots, plus urgent software-security fixes.
$5USD / month · plus applicable tax
Planned subscription: unlimited local scans and each tested detection release while subscribed.
Paid checkout is closed throughout the demo. These are future plan targets, not current entitlements. No automatic charge on December 1.
Compare access and update plansTell us what you connected, which warnings helped, and what got in the way. Leave out API keys and sensitive content.