Claude Managed Agents pricing calculator

Rate cards used, verified October 31, 2026

These prices are from the Anthropic pricing page, fetched October 31, 2026.

The pricing page quotes fast mode input at $8 and output at $40 USD per MTok. The fast mode cache prices are a calculation with the prompt caching multipliers, because the pricing page says that they apply on top of fast mode pricing. The 5m cache write is $8 x 1.25 = $10, the 1h cache write is $8 x 2 = $16 and the cache hit is $8 x 0.05 = $0.40 USD per MTok. Claude Haiku 5.5 prices each request by its own prompt length, and that length counts all of its input tokens, including cache reads and cache writes. Session totals do not show the prompt length of each request, so select the Haiku 5.5 level that matches your requests.

Session runtime replaces the code execution container-hour billing model when using Claude Managed Agents. You are not separately billed for container hours on top of session runtime. Batch API discounts do not apply because sessions are stateful and interactive, and there is no batch mode. Cloud platform pricing does not apply either, because Claude Managed Agents is not available on partner-operated cloud platforms such as Bedrock and Claude Models on Google Cloud (Google managed).

Worked examples from the official docs

Anthropic's pricing page gives two worked examples for a 1-hour coding session using Claude Opus 5.5 that consumes 50,000 input tokens and 15,000 output tokens.

• Without caching. Input 50,000 x $4 per MTok = $0.20 USD. Output 15,000 x $20 per MTok = $0.30 USD. Runtime 1.0 hour x $0.08 = $0.08 USD. Total $0.58 USD.
• With prompt caching, where 40,000 of the 50,000 input tokens are cache reads. Uncached input 10,000 x $4 per MTok = $0.04 USD. Cache reads 40,000 x $4 x 0.05 per MTok = $0.008 USD. Output $0.30 USD. Runtime $0.08 USD. Total $0.428 USD.

The default inputs above match the first example, so the calculator reproduces the official $0.58 USD figure when you load the page.

Session budget and max_list_cost

A session budget is an optional hard spend ceiling. A budget can only be attached when the session is created. The max_list_cost amount is a whole number of US cents written as a string with no leading zeros, so "125" is $1.25 USD and "50" is 50 cents. Decimal forms such as "25.00" are rejected. Enter a dollar cap in the calculator to get the amount string.

The budget compares against list cost, which is model tokens at list prices, web searches at $10 USD per 1,000 searches and session running time at $0.08 USD per hour. List cost is not your contracted price. With negotiated discounts, your billed spend can be lower than the cap. The cap is checked between model requests, so model spend past the cap is at most the cost of one model request for each thread. The tool calls in that request's response still run, and their web searches and running time also count toward list cost, so size the cap with that margin.

Fast mode, if speed matters more than cost

Fast mode provides faster output for Claude Opus 5.5 at premium token pricing, $8 USD per MTok input and $40 USD per MTok output instead of $4 and $20. Prompt caching and data residency multipliers stack on top of fast mode pricing. Fast mode applies when the agent's model.speed is set to "fast", and it is available on Claude Opus 5.5 for Managed Agents. The pricing page lists fast mode on the Claude API and, for Claude Opus 5.5, on Claude Models on Google Cloud (Anthropic managed) and Claude in Microsoft Foundry. It does not list Claude Platform on AWS. Select "Claude Opus 5.5, fast mode" in the calculator to compare it with a standard run at the same token volumes.

Ways to reduce session cost

• Cache aggressively. On Opus 5.5, cache reads cost 5 percent of the base input rate, $0.20 versus $4 per MTok. The official caching example cuts a 1-hour session from $0.58 to $0.428 USD.
• Watch the running clock. Runtime accrues only while the session status is running. Idle waits, rescheduling, and termination do not count, but a session left running while it thinks through a long task does. Shorter running time is a direct saving at $0.08 USD per session-hour.
• Cap sessions with a budget. You can set a hard list-cost budget at session creation. The platform stops issuing new model requests once the session reaches it.
• Pick the smallest model that passes your quality bar. Sonnet 5.5 output is $10 per MTok against Opus 5.5 at $20 per MTok.

Limitations

This calculator models the listed rates only. It does not model tool pricing beyond web search. On Claude Platform on AWS, session token and runtime charges convert to Claude Consumption Units at $0.01 USD per CCU. Claude Managed Agents is in beta on every model. LLM pricing shown is based on hardcoded reference values and may not reflect current provider rates. Always verify against official pricing pages before making budget decisions. The token counts you enter are estimates. Actual usage depends on encoding, model version, and prompt structure. This tool does not execute LLM calls. All computation happens locally in your browser.