TrustTest
Python red-teaming and evaluation framework that points built-in adversarial probes (prompt injection, data leakage, system-prompt extraction, agent and tool abuse) and functional tests at a model, agent or HTTP endpoint, generates test cases from datasets, RAG sources or prompts, and scores every response with heuristics and LLM-as-judge evaluators, locally or on the NeuralTrust platform.
Find alternatives to TrustTest
What it does
- Evaluation and observability
Sources
- Pricing
-
The pricing page prices TrustGate and TrustGuard only; it carries no TrustTest row, and the docs state no price.
- Description
-
Official documentation · 21 Oct 2026
"TrustTest is an AI red-teaming and evaluation toolkit. It runs adversarial attacks and functional tests against models, agents, and HTTP endpoints."
- What it does
-
Official documentation · 21 Oct 2026
"Adversarial attacks and functional quality checks in one framework — cleanly separating cases, evaluators and scenarios."
- Sold within
-
Docs: "Contact support if you do not have them yet." - the credential route is stated, a purchase route and price are not (pricing page https://neuraltrust.ai/pricing lists TrustGate and TrustGuard only).
- Name
-
Official documentation · 21 Oct 2026
Docs overview titled for the product: "TrustTest is an AI red-teaming and evaluation toolkit." The docs site is named "NeuralTrust Docs | TrustGate, TrustGuard, TrustTest".
- URL
-
Official documentation · 21 Oct 2026
Page of the product-footer entry "AI Red Teaming": "TrustTest is a red-teaming and evaluation framework that attacks your LLMs and agents with state-of-the-art adversarial techniques"; docs have a full TrustTest section (https://docs.neuraltrust.ai/trusttest/getting-started/overview).
- Deployment
-
Official documentation · 21 Oct 2026
FAQ: "Can results stay in my environment?" - local JSON results from the installed package, or the NeuralTrust platform ("Connect to NeuralTrust"); the red-teaming page (https://neuraltrust.ai/red-teaming) adds scheduled continuous runs on the platform.
- Interfaces
-
Official documentation · 21 Oct 2026
FAQ: "The Python package supports" HTTP and custom targets, and "The web UI currently requires an LLM exposed over REST." (a Python package and a web UI).
- Modalities
-
Official documentation · 21 Oct 2026
"Point it at a model, agent, or HTTP endpoint. Generate or supply test cases, score them, and keep the results" - prompts and probes go in as text; results are kept as JSON ("writes JSON locally") and in the platform.
- Intended for
-
Official documentation · 21 Oct 2026
"TrustTest is code-first." - define a target in code, version it in git and gate releases in CI; customer quote from a bank's transformation director.
Also from NeuralTrust
-
TrustGate Infrastructure service
TrustGate routes application and agent calls to models and tools while applying access controls, traffic policies, and guardrails.
-
TrustGuard Infrastructure service
Runtime security layer that inspects every request and response between AI agents and their models, MCP servers and tools across gateways, SDKs, browsers and sidecars, applying policies for prompt injection, PII and secrets, toxicity and topics with multi-turn context and an allow, block, transform, report or ask verdict on each interaction.
Open NeuralTrust in the directory
Something wrong here? Send a correction — quote this product id: neuraltrust-trusttest.