Stop a runaway agent by combining a hard limit on model calls or framework steps, cumulative per-run usage tracking, an enforced spend cutoff where your stack supports one, and alerts. Also cap each response’s output tokens. These controls do different jobs: a token budget shown to a model is not necessarily enforced, and a per-response limit does not stop an agent from making more requests.
Why one agent task can make many requests
An agent task can involve repeated model calls, tool executions, and handoffs before the application considers it finished. OpenAI’s Agents SDK describes a cycle in which the runtime calls the model, evaluates its output, executes tool calls or handoffs, and continues. Its runner can raise MaxTurnsExceeded when a configured max_turns limit is passed; setting max_turns=None disables that limit. See the OpenAI Agents SDK guide to running agents.
A costly run may involve repeated tool attempts, retries that compound a failure, or context that keeps growing and getting resent. Those are clues to investigate, not proof by themselves that a loop caused the bill. Inspect the trace and request-level usage before deciding what happened.
Put a hard ceiling on each run
Configure a maximum number of model calls, turns, or framework steps for each task. Choose the limit that corresponds to what your orchestration layer actually counts: a “turn” in one SDK may not equal a model request or a tool call in another. Settings and behavior can vary by framework version.
Recommended Free Tools
#1 Best Overall
- STYLISHLY SMALL, SLIM & DISCREET: Measuring just 3 1/8" x 4 7/16", our RFID front pocket wallet is designed to be super thin and exceptionally slim. Its modern, minimalist profile fits perfectly in your pocket, purse, or travel pack without adding bulk.
- SURPRISINGLY SPACIOUS: Though slim, it features 8 slots to easily organize your essentials. Comfortably holds your driver's license, credit cards, debit cards, and membership cards, keeping everything you need right at your fingertips.
- ADVANCED RFID BLOCKING: Our slim wallets for men and women are outfitted with advanced RFID SECURE Technology. They block electronic signals to keep your identity protected while you travel, shop, or explore, safeguarding you from digital theft.
- DURABLE & STYLISH FAUX LEATHER: Crafted from premium synthetic leather, this minimalist wallet sleeve combines a luxurious look and feel with everyday functionality. Its durable construction is designed to withstand the rigors of daily use, travel, and shopping.
- THE PERFECT UNISEX GIFT: With its sleek design and practical security features, this wallet is a popular choice for both men and women. It arrives ready for gifting, making it an ideal present for the frequent traveler, minimalist, or anyone in your life!
When the ceiling is reached, stop scheduling both model and tool work. Preserve the trace and return a clear status such as “incomplete—needs review,” ideally with any safe partial result. Do not silently restart the task as a fresh, uncapped run; that can erase the protection the limit was meant to provide.
LangChain’s current documentation describes model-call-limit middleware with run and thread limits. Check the API for the version you deploy in the LangChain model call limits guide.
Track cumulative usage and gate the next call
Record usage across every model request in the run, including requests that lead to tool calls or handoffs. Keep both a cumulative total and the individual request entries. A total helps enforce a run budget; the entries help identify which step, retry, or context expansion drove usage.
Rank #2
- Ultra-thin: This wallet measures 4.3 x 3 x 0.5 inches and can hold at least 11 cards and 15-20 bills. Even when it's packed full, it's only 0.8 inches thick,It can perfectly conceal itself in your pocket without any noticeable bulge.
- Rfid Blocking: Our wallets are equipped with German Instiute Certified RFID Security technology, a unique metal composite, engineered specifically to block 13.56 MHz or higher RFID signals to protect the valuable information and privac.
- Lifetime After-sales Service: Regardless of the circumstances, if any GSOIAX brand wallet has a quality issue during your use, we promise to provide a full, unconditional, refund within 24 hours!
- Durable Surface: Crafted from premium 3-layer leather, our wallets outperform 2-layer alternatives in durability. Specially treated leather exterior delivers enhanced scratch resistance to guard against minor scuffs from everyday items like keys and buttons.
- Perfect Gifts For Him: This Money Clips Wallets for men comes in classy gift box package. It's a good idea to send the mens wallets as the gifts in birthday，anniversaries, Fathers Day，Valentine's Day，Christmas and other special occasions to someone you love.
The OpenAI Agents SDK exposes request counts, input and output tokens, totals, and per-request usage entries. Its reporting can vary with third-party provider adapters and some streaming backends, so verify that the exact adapter and execution path you deploy report what you expect. See OpenAI Agents SDK usage tracking.
Before issuing the next model request—or an expensive tool operation—check the run’s remaining allowance. If the framework does not offer an enforced run cap, use a wrapper or gateway only after verifying that it actually blocks the next operation at your chosen threshold. An alert or a usage counter is not a stop mechanism.
Know which budget actually stops work
Limits can apply at different scopes and enforce different things. A model-facing budget may encourage an agent to finish sooner; a per-request output cap constrains one response; a runtime ceiling ends a run; and a dollar limit may reject further spending within a managed service. Do not treat these as interchangeable.
Rank #3
- RFID Blocking Technology: This credit card holder is made of aluminum shells and ABS plastic, designed with RFID-blocking technology to help protect your credit, ID, debit, and driver's license cards from unauthorized scanning
- Slim Compact: Slim and compact design measures 4.3 x 3 x 0.86 inches, ideal for front pockets or purses
- Card Organizer: With 7 accordion-style slots, this wallet can hold up to 10 standard credit cards or over 20 business cards
- Artistic Expression: Features a variety of artistic designs on the aluminum shell, inspired by famous paintings, flowers, and animals, to complement your personal style
- Thoughtful Gift Idea: Makes a thoughtful gift for any occasion, combining functionality and style
| Control | Scope and enforcement | What it does not guarantee |
|---|---|---|
| Loop-level token budget | Advisory guidance to the model during a task, where supported. | It may be exceeded and is not necessarily an enforced dollar cap. |
| Per-request output limit | Caps generated output for one model request. | It does not limit how many requests the loop can make. |
| Run or step ceiling | Terminates or blocks further work after the configured count. | It does not automatically account for every dollar or tool charge. |
| Session or workspace spend limit | Can act as a dollar cutoff at the scope supported by the managed product. | Coverage, availability, and accounting vary; confirm whether retries, tools, or nested agents count. |
Anthropic task budgets and output limits
Anthropic describes task budgets as a soft hint: the model may exceed the token countdown to finish an action, and the budget is not returned in the response usage object. Its documentation distinguishes this from max_tokens, the enforced per-request output ceiling. Pair a per-request ceiling with a loop limit if you need to constrain repeated calls. Read Anthropic’s agentic loops documentation.
Managed-service dollar limits
For supported managed-agent workflows, Anthropic’s cost guidance describes a session budget as a hard dollar stop and a workspace spend limit as a broader backstop. AWS likewise recommends layered cost controls, including per-cycle, per-task, and per-day limits with automatic cutoffs. Product, model, and regional availability can change; verify the behavior in your own account and deployment before relying on a setting as a hard stop. See Anthropic cost and pricing guidance and AWS agentic AI cost optimization guidance.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchProvider account-level spending controls do not have one universal behavior across vendors and plans. Confirm whether a limit rejects new requests, merely sends an alert, or applies only after usage is reported or reconciled.
Rank #4
- SECURE YOUR WALLET FROM e-PICKPOCKETING: Prevent potential identity and financial theft through your contactless cards. This is the simplest and most effective prevention solution! Block RFID and NFC signals, protect your personal information, and enjoy peace of mind wherever your travels or business take you.
- JAMMING CHIP: An antenna and jamming chip makes up the main components of the card. The antenna will sense incoming radio waves and draw power for the chip to create a jamming signal. Lifetime usage as the card does not require battery.
- BROAD WORKING DISTANCE: With a 2.4” working distance, your entire wallet stays protected. The premium RFID blocking card helps secure cards within 1.2” on either side, providing reliable protection against electronic pickpocketing.
- ULTRA-THIN & COMPACT: At the size of a standard credit card and at only 0.03” thick, the card will fit into any wallet, purse or card case. Keep your wallet compact with no added bulk from this card. Best for travel, business, and everyday use.
- TEST THE CARD: Test the card is working at your local supermarket. At the self-service checkout machines, combine the card and a contactless card on the payment reader. Payment with the contactless card will be blocked and an error message should occur on the reader.
Choose thresholds from real task data
There is no universal safe number of calls or tokens for every agent. A short retrieval task and a multi-step coding task can have different normal usage. Start with representative tasks, then set distinct ceilings where task classes materially differ.
- Measure a baseline. Record model calls, input and output usage, estimated or reconciled spend, tool calls, completion status, and latency by task type.
- Set an initial ceiling. Use successful runs to establish a practical range. Anthropic suggests using observed high-percentile usage, such as p99, when tuning its advisory task-budget feature; that is vendor guidance for that feature, not a universal dollar formula.
- Test failure cases. Exercise tool timeouts, repeated tool errors, retries, handoffs, and unusually large context. Confirm the limit stops future work and produces a usable failure status.
- Roll out gradually. Review capped runs, quality, and completion rates, then adjust thresholds without removing the hard outer safety limit.
Compare cost per completed or accepted task, not just total spend. A cap can make the bill smaller by aborting useful tasks too early; include completion quality and latency when deciding whether the threshold is working. AWS recommends outcome-oriented measures such as cost per task completion alongside cost controls.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Monitor for escalation and prepare a response
Track signals at the run or agent identity level so an individual runaway task does not disappear inside an account-wide average. Useful metrics include:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsBest Value
- Special Design: Multi-color optional and wear-proof classic business card holder looking.
- Plenty of Space: 16 card slots only measuring 4.1" x 3.0" x 1.1", including 13 credit card slots, 2 cash slots
- Protect Information Leakage: Prevents your vital information/cards from unnoticed scan with 2 outer layers RFID blocking materials.
- Extra Key Chain & Portable: Extra corns with key chain for your keys or lanyard. Portable use for shopping, traveling, etc.
- Great Gift: Practical compact wallet is the perfect gift. Give a thoughtful surprise to Men/Women on birthdays, holidays, celebrations, or any special occasion (e.g. Valentine's Day, Christmas, etc.).
- Model request count and input/output usage per run.
- Tool invocations, retries, and handoffs per task.
- Estimated spend or reconciled charges, where available.
- Context or memory growth over the run.
- Completion state and the reason the run stopped.
Set alerts for unusual rates and threshold crossings, and send the operator the trace and stop reason. AWS calls out token spikes, tool-invocation storms, and memory growth as agent-specific signals worth monitoring. An alert helps detect an incident; only a control that blocks or terminates further work limits additional exposure.
Reduce ordinary run costs after bounding the loop
Once a runaway run is bounded, optimize the work that completes normally. Repeated prompt context can add cost across many turns; caching repeated context or trimming unnecessary input may reduce that cost, but neither measure stops the agent from continuing.
Anthropic reports a 2.7–5.3× reduction in agent-loop cost from prompt caching on benchmarks described in its 2026 cost guide. In a documented small triage-agent test, caching reduced the bill by 83%, or by 88% when input trimming was added. These are vendor-published results for those workloads, not guaranteed savings for another agent. The same guide describes a measured run in which context editing cost more than it saved. Measure cache-hit patterns, actual costs, and task quality in your own system before keeping an optimization. See Anthropic cost and pricing guidance.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →
