[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-agent-budget::en":3,"gloss-cluster-agent-budget::en":26,"gloss-next-agent-budget::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"agent-budget","agents","Agent Budget","An agent budget is a hard cap on how much an agent may consume before it must stop or escalate — measured in steps (loop iterations), tokens, tool calls, wall-clock time, or dollars. Because an agent decides its own next action, a bad plan or a loop of two agents talking past each other can spin indefinitely, quietly burning money and API rate limits; a budget is the circuit breaker that prevents a runaway run. When the limit is hit, the agent stops, returns partial results, or hands off to a human rather than grinding on. For builders, budgets are non-negotiable guardrails the moment an agent has real autonomy or touches paid APIs: set a max-steps ceiling to catch infinite loops, a token or cost ceiling to bound spend per task, and log usage so you can tune later. Pair budgets with clear stop conditions so hitting the cap is a graceful exit, not a silent failure.","An agent budget is a hard cap on steps, tokens, tool calls, time, or dollars — because an agent picks its own next action and a bad plan can spin indefinitely.",null,[11,14,17,20,23],{"slug":12,"name":13},"agent-loop","Agent Loop",{"slug":15,"name":16},"autonomous-agent","Autonomous Agent",{"slug":18,"name":19},"guardrails","Guardrails",{"slug":21,"name":22},"human-on-the-loop","Human-on-the-Loop",{"slug":24,"name":25},"latency","Latency",[27,31,35,36,39,42,45,48,51,54,57,61],{"slug":28,"category":5,"name":29,"updated_at":30},"agent-checkpointing","Agent Checkpointing","2026-08-24T02:46:38+00:00",{"slug":32,"category":5,"name":33,"updated_at":34},"agent-handoff","Agent Handoff","2026-08-24T02:46:37+00:00",{"slug":12,"category":5,"name":13,"updated_at":34},{"slug":37,"category":5,"name":38,"updated_at":34},"agent-memory","Agent Memory",{"slug":40,"category":5,"name":41,"updated_at":30},"agent-sandbox","Agent Sandbox",{"slug":43,"category":5,"name":44,"updated_at":30},"agent-trajectory","Agent Trajectory",{"slug":46,"category":5,"name":47,"updated_at":34},"agentic-rag","Agentic RAG",{"slug":49,"category":5,"name":50,"updated_at":30},"computer-use","Computer Use",{"slug":52,"category":5,"name":53,"updated_at":34},"context-engineering","Context Engineering",{"slug":55,"category":5,"name":56,"updated_at":30},"dry-run","Dry Run",{"slug":58,"category":5,"name":59,"updated_at":60},"escalation-path","Escalation Path","2026-08-24T03:30:02+00:00",{"slug":21,"category":5,"name":22,"updated_at":34}]