Claude Managed Agents pricing, $0.08 per session-hour plus model tokens
Claude Managed Agents bills the tokens of a session at the model prices and the runtime of the session at $0.08 per session-hour. This page gives the two official worked examples, the modifiers that apply and a session cost calculator.
Each official price on this page is from one source, the Claude Platform pricing page of Anthropic, fetched on 10 October 2026 at 08:21 UTC. Prices are USD list prices. They can change without notice. KickLLM is not affiliated with Anthropic.
Claude Managed Agents is billed on two dimensions: tokens and session runtime. All tokens that a session consumes are billed at the prices of the Model pricing table. Prompt caching multipliers apply identically. A web search triggered inside a session costs the standard $10 per 1,000 searches. Session runtime is $0.08 per session-hour, and it accrues only while the status of the session is running.
Two Messages API modifiers do not apply to a Managed Agents session. The Batch API discount does not apply, because sessions are stateful and interactive and there is no batch mode. Cloud platform pricing does not apply, because Claude Managed Agents is not available on partner-operated cloud platforms. Session runtime also replaces the code execution container-hour billing model, so you are not separately billed for container hours on top of session runtime.
The two billing dimensions
Claude Managed Agents, USD, from the Claude Platform pricing page fetched on 10 October 2026
| Dimension | Price | How it is metered |
|---|---|---|
| Tokens | The prices of the Model pricing table | All tokens that the session consumes |
| Session runtime | $0.08 per session-hour | Duration of the running status |
Runtime is measured to the millisecond. It accrues only while the status of the session is running. Time in the idle, rescheduling or terminated status does not count toward runtime. In the idle status the session waits for your next message or for a tool confirmation.
On Claude Platform on AWS, session token and runtime charges convert to Claude Consumption Units at the standard rate.
Which modifiers apply inside a session
Messages API modifiers and Claude Managed Agents sessions, same source and date
| Modifier | Applies to a session? | What the pricing page says |
|---|---|---|
| Prompt caching multipliers | Yes | They apply identically |
| Web search | Yes | The standard $10 per 1,000 searches, for a search triggered inside a session |
| Fast mode premium pricing | Yes | It applies when model.speed of the agent is set to "fast" |
| Data residency multiplier | Yes | Tokens are billed at 1.1x the standard rates when model.inference_geo of the agent is pinned to "us" |
| Batch API discount | No | Sessions are stateful and interactive. There is no batch mode. |
| Cloud platform pricing | No | Not available on partner-operated cloud platforms |
The pricing page names Amazon Bedrock and Google Cloud as the partner-operated cloud platforms.
The two official worked examples
Both examples are from the Claude Platform pricing page fetched on 10 October 2026. Each one is a one-hour coding session on Claude Opus 5 with 50,000 input tokens and 15,000 output tokens.
Official example 1, no cache reads, total $0.705
| Line item | Calculation | Cost |
|---|---|---|
| Input tokens | 50,000 x $5 / 1,000,000 | $0.25 |
| Output tokens | 15,000 x $25 / 1,000,000 | $0.375 |
| Session runtime | 1.0 hour x $0.08 | $0.08 |
| Total | $0.705 |
Official example 2, 40,000 cache-read tokens, total $0.525
In the second example prompt caching is active, and 40,000 of the 50,000 input tokens are cache reads.
| Line item | Calculation | Cost |
|---|---|---|
| Uncached input tokens | 10,000 x $5 / 1,000,000 | $0.05 |
| Cache read tokens | 40,000 x $5 x 0.1 / 1,000,000 | $0.02 |
| Output tokens | 15,000 x $25 / 1,000,000 | $0.375 |
| Session runtime | 1.0 hour x $0.08 | $0.08 |
| Total | $0.525 |
Session cost calculator
KickLLM estimate. Prices are from the Claude Platform pricing page fetched on 10 October 2026
| Line item | KickLLM estimate |
|---|---|
| Tokens | $0.625 |
| Session runtime | $0.08 |
| Web search | $0.00 |
| Total | $0.705 |
Each output is a KickLLM estimate. It is not an Anthropic figure, and your invoice can be different. The start values are official example 1. Set the cache-read tokens to 40,000 to get official example 2. Cache-read tokens are a part of the input tokens. The calculator applies the US-only multiplier of 1.1x to the token cost only. It does not include cache writes or fast mode. A web search is $10 per 1,000 searches. The model list is six rows of the general Model pricing table of the pricing page. It is not a list of the models that Claude Managed Agents supports. Each amount is rounded to 6 decimal places at most.
Prices that the calculator uses, USD per 1M tokens, from the Model pricing table of the same page, fetched on 10 October 2026
| Model | Base input | Cache read | Output |
|---|---|---|---|
| Claude Opus 5.5 | $4.00 | $0.20 | $20.00 |
| Claude Opus 5 | $5.00 | $0.50 | $25.00 |
| Claude Sonnet 5.5 | $2.00 | $0.10 | $10.00 |
| Claude Sonnet 5 | $2.00 | $0.20 | $10.00 |
| Claude Haiku 5.5 (for prompts up to 100,000 tokens) | $0.10 | $0.01 | $0.50 |
| Claude Haiku 5.5 (for prompts over 100,000 tokens) | $0.50 | $0.05 | $2.50 |
The pricing page does not list the models that Claude Managed Agents supports. The calculator uses six rows of its general Model pricing table, which shows the prices of all Claude models. Claude Haiku 5.5 has two rows. A request whose prompt is over 100,000 tokens pays the higher prices, and each request is priced on its own. Select the row that agrees with your requests.
Session runtime and container hours
The pricing page says that the code execution tool is billed by execution time when it is used without web search or web fetch. It gives each organization 1,550 free hours per month. Usage above that is $0.05 per hour, per container. In Claude Managed Agents, session runtime replaces this container-hour billing model. You are not separately billed for container hours on top of session runtime.
How these numbers were checked
There is one source: the Claude Platform pricing page, fetched as text on 10 October 2026 at 08:21 UTC. Its section "Claude Managed Agents pricing" gives the two billing dimensions, the $0.08 per session-hour rate, the modifiers and the two worked examples with the totals $0.705 and $0.525. The model prices in the calculator are from the Model pricing table of the same page. Each Anthropic price on this page comes from that one official page, as it was at that fetch time. No second source was used.
The calculator outputs are KickLLM estimates. They are not Anthropic figures. The prices are USD list prices of the fetch date, and they can change without notice. KickLLM is not affiliated with Anthropic.
Frequently asked questions
How much does a Claude Managed Agents session cost per hour?
Session runtime is $0.08 per session-hour, and it accrues only while the status of the session is running. Tokens are billed in addition, at the model prices. Thirty minutes in the running status is $0.04 of runtime (0.5 x $0.08, a KickLLM calculation).
Do prompt caching and web search apply inside a Managed Agents session?
Yes. Prompt caching multipliers apply identically, and a web search triggered inside a session costs the standard $10 per 1,000 searches. The Batch API discount and cloud platform pricing do not apply to Managed Agents sessions.
Is idle time billed as session runtime?
No. The pricing page says that time in the idle, rescheduling or terminated status does not count toward runtime. Runtime is measured to the millisecond, in the running status only.
Is Claude Managed Agents billed for container hours?
No. Session runtime replaces the code execution container-hour billing model. You are not separately billed for container hours on top of session runtime.