A retry budget is a hard limit on the number of retry attempts an agent can make before escalating or failing gracefully. Delx recommends a default budget of 3 retries with exponential backoff (1s, 2s, 4s). Exceeding the budget triggers recovery instead of further retries.
Retry budgets prevent retry storms — a cascade of retries that overwhelm downstream services and create more failures than they resolve. Without a budget, an agent hitting a 503 might retry hundreds of times in seconds, making the outage worse. The budget pattern works in 3 layers: (1) per-call budget — maximum retries for a single tool call (default: 3), (2) per-session budget — total retries across all calls in a session (default: 15), (3) per-window budget — retries allowed in a time window, typically 10 per minute. When any budget is exhausted, the agent should call process_failure to log the pattern and recovery to get alternative strategies. Backoff strategy matters too: exponential backoff (doubling wait time) is standard, jittered backoff (random variation) prevents thundering herd, and circuit breaker (stop retrying for N minutes) protects degraded dependencies.
A production agent deployment uses retry budget to improve reliability and observability. The pattern is standard across MCP, A2A, and REST protocols.
When investigating agent failures, retry budget provides structured data that helps identify root causes and track resolution progress.
A retry budget is a hard limit on the number of retry attempts an agent can make before escalating or failing gracefully. Delx recommends a default budget of 3 retries with exponential backoff (1s, 2s
Retry Budget 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 retry budget through its MCP and A2A protocol endpoints. The pattern is consistent across all Delx tools and available via REST, CLI, and direct protocol calls.