Turn a mission need into explicit objectives, success metrics, failure conditions, evidence requirements, and review gates.
AVAILABLEFEDERAL EVALUATION MODE · PUBLIC BETA v0.1
Define the mission. Then test the evidence.
Build an evaluation plan for an AI system, agent, commercial product, dataset, or digital service. The plan freezes the objective, acceptance criteria, evidence, independent review, and monitoring before results are known.
Public-beta boundary: use public or synthetic information only. Do not upload classified information, CUI, PII, PHI, procurement-sensitive material, credentials, export-controlled information, or other protected data. This public beta is not FedRAMP-authorized and is not an ATO environment.
CAPABILITY DECLARATION
What Mission Control can actually do today.
Create mission-context test packs for AI/agents, software/COTS, data, and digital services; preserve negative and UNKNOWN outcomes.
PILOTRoute bounded public/synthetic evaluation toward Testing Ground and provenance-first Great Library records.
PILOTGenerate performance criteria suitable for product demonstrations, acceptance tests, QASP-style monitoring, and post-award re-test planning.
PILOTScope accessibility criteria and require evidence, manual review, re-test, and remediation tracking.
PLAN AVAILABLE · EXECUTION PARTIALDefine attack/failure hypotheses, software-security evidence, regression tests, and independent review. Production probing is not authorized by default.
PLAN AVAILABLE · EXECUTION BOUNDEDSpecify monitoring metrics, reassessment cadence, significant-change triggers, rollback/fallback, and evidence retention.
PILOTThe public beta cannot receive protected federal workloads. These require an approved isolated environment and the agency's authorization process.
NOT AUTHORIZED HERE1 · SYSTEM
2 · STAGE
3 · CRITERIA
4 · PLAN
What are you evaluating?
Pick the closest category. The test plan stays technology-neutral.
Describe the decision or mission outcome—not a preferred vendor.
FEDERAL BASIS
Why these test lanes exist.
High-impact AI requires pre-deployment testing, impact assessment, independent review, and lifecycle reassessment tied to real-world outcomes.
Official memo →Agencies are encouraged to test products in realistic scenarios, use performance-based requirements, consider QASP-style monitoring, interoperability, and vendor lock-in.
Official memo →Customized TEVV for real organizational objectives across statistical ML, LLMs, multimodal models, and agentic systems.
Official NIST framework →Federal agencies need rigorous testing and evaluation to determine whether commercial products meet operational requirements before purchase.
Official GAO guide →Federal ICT needs repeatable accessibility testing throughout development, procurement, acceptance, and maintenance; automated testing alone is insufficient.
Official Section 508 testing →DHS has called for independent assessment, benchmarking, validation, monitoring, acquisition testing, AI testbeds, and security research for AI-enabled systems.
Official DHS roadmap →