toolgrep / AI agents & frameworks
AgenticRail
Deterministic sequence-enforcement API for AI agents that returns ALLOW, DENY, or HALT before each step and issues an Ed25519-signed, hash-chained receipt for every decision.
Visit agenticrail.nz →DocsPricing
AgenticRail is a runtime sequence-enforcement gate for AI agents and other automated processes. Callers POST each proposed step to /v1/evaluate with a sequence_id, step, action_type, nonce, and optional timestamp; the gate validates step order against a declared step_order, checks nonce replay, enforces action_type policy, and confirms the sequence is not sealed. Permitted steps return ALLOW with a signed receipt; violations return DENY with reason codes such as SEQUENCE_VIOLATION, REPLAY_NONCE, STALE_TIMESTAMP, SEALED_SEQUENCE, and STEP_ORDER_MISMATCH. Receipts are Ed25519-signed, hash-chained, and verifiable offline via a compliance report service. The architecture consists of a public wrapper (api.agenticrail.nz), an internal gate, and a read-only report service (report.agenticrail.nz). MIT-licensed Python and JavaScript SDKs ship LangGraph and CrewAI integrations, and an MCP server exposes evaluate_step and verify_receipt tools. The underlying protocol is SLP-8.
- Category
- AI agents & frameworks
- Pricing
- freemium · from US$39/month
- Free tier
- yes
- API
- yes
- MCP server
- yes
- SDK
- yes
- Open source
- yes
Listed 2026-10-29 · last verified 2026-10-29 · extracted by @cf/zai-org/glm-5.3-flash · confidence 0.9
Your product? Claim this listing · Report a problem · Badge
API
curl https://toolgrep.com/api/products/agenticrail