Action State launches tools to verify agent interactions

Action State Group Inc. is launching open-source tools to help businesses independently check records of AI agents buying, selling and delegating work across corporate boundaries. The offering combines EvidenceBook for tamper-evident records, Terms Packs for evaluating behavior and Countersign for independent verification.

EvidenceBook gives participants signed transaction histories they can selectively compare to identify agreements, conflicts or gaps while retaining control of their data. Verification is free and can occur offline. Signed records and cryptographic checkpoints make subsequent changes detectable, but preserving evidence does not establish that its contents are accurate or complete. Countersign checks cryptographic and structural properties, signs its findings and logs them without certifying claimed outcomes.

Terms Packs evaluate recorded work against business or regulatory requirements, identifying conditions that were met, missed or could not be assessed. An initial pack covers agent-related obligations under the European Union’s AI Act, but does not certify legal compliance. Action State recommends customization by governance teams and human review of automated judgments.

Early integrations include Mesh-LLM and Project Nanda, while Everyday Rules connects with Muse to evaluate personal agents’ proposed actions against owner requirements. EvidenceBook uses the Apache 2.0 license. Action State’s specifications remain individual Internet-Drafts rather than adopted standards. The company plans to seek neutral foundation governance and eventually charge for Countersign, which is currently free.