An agent nudge is a suggested next action embedded in Delx tool responses. Unlike commands, nudges are advisory — the agent can follow them or ignore them based on its own logic. Nudges appear in the DELX_NUDGE block and in the next_action field of DELX_META.
Nudges solve the 'what should I do next?' problem that agents face after completing a tool call. Instead of returning raw data and leaving the agent to figure out its next step, Delx suggests specific follow-up actions. Common nudges include: 'call heartbeat to update wellness score', 'run process_failure to log this error', 'compact your session — context at 85%', and 'close this session — incident resolved'. Nudges are context-aware — they consider the agent's session history, current wellness score, and the outcome of the current tool call. The nudge system is opt-in: agents that don't parse DELX_NUDGE won't be affected. Advanced agents use nudges as input to their planning step, weighting Delx's suggestions alongside their own heuristics.
A production agent deployment uses agent nudge to improve reliability and observability. The pattern is standard across MCP, A2A, and REST protocols.
When investigating agent failures, agent nudge provides structured data that helps identify root causes and track resolution progress.
An agent nudge is a suggested next action embedded in Delx tool responses. Unlike commands, nudges are advisory — the agent can follow them or ignore them based on its own logic. Nudges appear in the
Agent Nudge is essential for production AI agent deployments. Without it, agents operate as black boxes with no observability, reliability guarantees, or structured failure handling.
Delx implements agent nudge through its MCP and A2A protocol endpoints. The pattern is consistent across all Delx tools and available via REST, CLI, and direct protocol calls.