Stops agents double-firing side effects like double-charges or duplicate sends: same-instant races elect exactly one winner, and late duplicates get a sealed, content-addressed receipt replayed instead of a second execution. Tools: fence_prepare, fence_commit, fence_abort.
Provides durable mission state for AI agents, enabling crash-safe recovery, verified completion, and explicit reconciliation of uncertain side effects without duplicate or blind retries.
Universal verifiable recovery for long-running AI agents with semantic checkpoints, idempotent action ledger and hash chained log as a deny by default MCP server. Framework agnostic with adapters for LangGraph, LangChain and OpenAI plus gateway and OTel.
Deterministic execution engine for AI agents. 412 modules across 78 categories including browser automation, file I/O, Docker, data parsing, crypto, and scheduling. Supports
STDIO and Streamable HTTP transport with execution trace, evidence snapshots, and replay from any step.
Safe, reversible tool execution for AI agents. It sits between an agent and its tool servers, adding contracts, dry-run planning, policy, approvals, saga execution, and rewind.
Enables AI agents to gate real-world side effects through a durable decision record, ensuring at-most-once initiation, deduplication, budget enforcement, and auditable outcomes across retries and failures.
