Stateful by Design, and Documented as Such — Luminity Digital
The Published Boundary — Where the Harness Ends  ·  Series 32  ·  Post 2 of 4  ·  August 2026
Claude Architect Dispatches

Stateful by Design, and Documented as Such

Anthropic publishes the constraint most consequential for regulated work — and the same statefulness that produces it is what makes the arc’s only currency-denominated ceiling and its only pausing, reason-returning tool gate possible.

August 2026 Tom M. Gomez Luminity Digital 12 Min Read
Third post in The Published Boundary — Where the Harness Ends, and a re-read of The Session Is the Substrate on a narrower axis. That dispatch read Claude Managed Agents as a write-path surface. This one asks the arc’s three questions: where the session record resides, what the hard limit is denominated in, and where the tool gate is declared. Companion dispatches read AgentCore harness and OpenAI Presence against the same three, and the closeout tabulates all three. Every claim here is sourced to Anthropic documentation.

The Shared Responsibility Model Comes to the Harness found that Anthropic named what is owned at each layer without documenting the seam.

For Managed Agents specifically, that is no longer the case, and the documentation is unusually direct about the thing a risk committee will ask first.

Anthropic publishes that Managed Agents carries neither zero-retention nor healthcare-agreement eligibility, states the reason — sessions are stateful resources whose transcripts persist until deleted — and extends the statement to every sub-feature of the product, self-hosted sandboxes included [5]. That is a vendor documenting a constraint on its own product in the place an architect checks before assigning a workload. It is the seam, written down.

The rest of this dispatch reads what that statefulness buys, because the same property that produces the eligibility statement produces the strongest cost control and the strongest approval gate in this arc.

Five Bands, One Held Together

Canonical harness frame annotated for Managed Agents responsibilityThe five bands with responsibility per band on Claude Managed Agents.Caller and authenticationApplication key, caller policy above itCustomerAgent definitionModel, prompt, toolsets, permission policiesCustomerOrchestration loopLoop and session record held abovePlatformExecution sandboxManaged by default, self-hosting electivePlatformTools and model providersGate above, toolset and scope belowSharedCustomerPlatformSharedThe orchestration loop and the session record are held together, as one band.Electing a self-hosted sandbox moves the fourth band without moving the third.
Figure 01  ·  Canonical Frame, Managed Agents Responsibility Annotations The orchestration loop and the session record are a single platform-held band — the one place this frame differs from AgentCore. The top two bands sit with the customer; the tools band is shared. Scroll figure horizontally to view in full

Compare this to the AgentCore frame and one band carries the whole difference. There, the orchestration loop is the loop. Here, the loop and the session record are one band, held together on the platform side. Anthropic runs the loop and keeps the record of what the loop did.

The top two bands sit with the customer, as they do on the other platform: the agent definition — model, system prompt, toolsets, permission policies — is authored by the customer, and authorization of callers into the customer’s own application is the customer’s concern. The tools band is shared, because the platform supplies the gate and the customer declares which toolsets exist and what they reach.

Where Does the Session Record Reside?

With Anthropic, and electing a self-hosted sandbox does not change that.

Control plane and execution plane under both sandbox electionsSelf-hosting relocates tool execution while orchestration and the session record remain platform-side.Managed sandboxControl planeOrchestration loop, model calls,session recordExecution planeFilesystem, shell, tool runsControl planeOrchestration loop, model calls,session recordExecution planeCustomer infrastructureSelf-hosted sandbox, by electionPlatformCustomerTool inputs and outputs still reach the control plane, so the model can act on results.The election moves where tools run. It does not move where the session record lives.
Figure 02  ·  Control Plane and Execution Plane Under Both Sandbox Elections Electing a self-hosted sandbox relocates tool execution. Orchestration stays on Anthropic’s side and tool inputs and outputs still reach the control plane, because the model has to see results in order to act on them. Scroll figure horizontally to view in full

This is the distinction most worth internalizing before a compliance review. Anthropic offers a managed cloud sandbox and a self-hosted sandbox running on the customer’s own infrastructure [1]. Electing the latter relocates tool execution. Orchestration remains on Anthropic’s side, and tool inputs and outputs still reach Anthropic’s control plane, because the model has to see results in order to act on them [4]. The execution boundary and the record boundary move independently, and only the second one settles the processor arrangement.

Session lifecycle, persistence, and deletionA session persists as a stateful resource until deleted through the API.CreatedActiveIdleAwaiting a decisionPersists as a stateful resourceEvent history, sandbox state, and outputs remain retrievable until removedDeleted through the APISessions and uploaded files are removable on the customer’s scheduleStatefulness is the design, and Anthropic publishes what follows from it: the product carries nozero-retention or healthcare-agreement eligibility, across every sub-feature including self-hosting.
Figure 03  ·  Session Lifecycle, Persistence, and Deletion Event history, sandbox state, and outputs persist server-side and remain retrievable until removed. Deletion through the API makes retention a schedule the customer operates rather than a fixed window. Scroll figure horizontally to view in full

Sessions persist event history, sandbox state, and outputs server-side, and remain retrievable until removed [1]. Deletion is available through the API, so retention is a schedule the customer can operate rather than a fixed window [1]. What is not available is a commitment that the record was never retained at all, which is exactly what the eligibility statement says, and why it says it [5].

One further property belongs here, because it is a read-path control that is easy to mistake for a write-path one. An agent can pin an inference geography, and Anthropic enforces coherence across a multiagent roster: the coordinator’s pin and every member’s pin must all be set to the same value or all be unset, with a mismatch rejected as a validation error at save time and at session-create override [6]. That is genuine residency control, enforced structurally rather than by convention. It governs where inference happens. It does not change who processes the session record.

What Is the Hard Limit Denominated In?

Currency. This is the only currency-denominated ceiling in the arc, and it is the strongest single assurance primitive across all three platforms.

What a session accepts and refuses once its budget is reachedSettling events are accepted; events that would start new work are refused. The session idles.Ceiling reachedPriced continuously at list rates, evaluated between model requests, so overshoot is boundedAccepted — settles work already in flightTool resultTool confirmation decisionInterruptRefused — would start new workA new message from the userThe session goes idle rather than terminating. Conversation history and sandbox state are preserved.Raising the ceiling resumes it. Removing the ceiling entirely cannot be undone.
Figure 04  ·  What a Session Accepts and Refuses Once Its Budget Is Reached At the ceiling the session goes idle rather than terminating. Events that settle work already in flight are accepted; an event that would start new work is refused. Scroll figure horizontally to view in full

A session can carry an optional cap in whole cents, priced continuously at public list rates and evaluated between model requests, so overshoot is bounded by at most one request per thread [3]. At the ceiling the session goes idle rather than terminating, preserving conversation history and sandbox state [3]. Events that settle work already in flight — a tool result, a tool confirmation decision, an interrupt — are accepted. An event that would start new work is refused [3]. That is a control that fails closed on new spend while remaining open to completing what it already started, which is the correct shape for a long-running agent.

In a multiagent session the cap is a single shared ceiling across all threads, threads pause independently as it is reached, and each thread’s cost is priced at the model that thread was actually served by [6]. A budget therefore survives the case a token ceiling cannot: work distributed across several models still resolves to one number in one unit.

Two conditions belong in any specification that relies on it. Removing a budget from a session cannot be undone [3]. And a model without a public list price cannot be priced continuously, so it cannot be budgeted [3]. Both are stated by Anthropic, and both are the kind of detail that determines whether a control holds in production or only in a diagram.

Where Is the Tool Gate Declared?

In a configuration object the customer authors, and it pauses execution.

Permission policy evaluation, pause, and the two decision pathsAn allow decision proceeds; an ask decision pauses the session pending a human decision.Model requests a toolwithin a toolsetPolicy evaluatedPer toolset, per tool overrideallowTool executesResult tothe modelaskSession pausesAwaits a decisionfor as long as it takesapproveddeniedDenial reason returned to the model,which continues with that knowledgeAgent toolsets default to allow.Connected server toolsets default to ask.
Figure 05  ·  Permission Policy Evaluation, Pause, and the Two Decision Paths An ask decision does not fail the call. The session goes idle awaiting a decision, and a denial returns a reason to the model, which continues with that knowledge rather than retrying blindly. Scroll figure horizontally to view in full

Permission policies are declared per toolset with per-tool override. Agent toolsets default to allow; connected server toolsets default to ask [2]. An ask decision does not fail the call — the session goes idle awaiting a decision and waits as long as it takes, and a denial returns a reason to the model, which continues with that knowledge rather than retrying blindly [2]. In a multiagent session the request is cross-posted to the primary thread with the originating thread identified, so a subagent’s permission request surfaces where a supervising client is already listening [6].

Three properties make this reviewable rather than merely present. The policy is authored, so a reviewer can read it and a release process can diff it. The default for connected servers is the conservative one, so a new external toolset arrives gated rather than open. And the denial is informative, so a refusal becomes part of the agent’s context rather than a silent dead end.

Change control exists here too. Agents are versioned, and a coordinator’s configuration — including its roster of delegable agents — is snapshotted when the coordinator is created or updated. Referenced agents stay pinned to the versions resolved at that moment and do not pick up later updates; delegating to a newer version requires updating the coordinator [6]. That is a pinning construct, and it means the composition of a multiagent session is a fixed artifact rather than a live lookup.

What the Customer Owes

Obligation 01  ·  Scope, Authored per Agent

Scope is declared per agent, and Anthropic documents it as the mechanism. MCP servers are agent-scoped while vault credentials are session-scoped, and the guidance for limiting an agent’s reach is to declare only the servers that agent needs [6]. That is least privilege expressed as configuration rather than as practice — and it is the customer’s to author. A session’s vault credentials reach every thread, so a credential supplied for one agent’s server is present in a session where other agents run.

Obligation 02  ·  Composition Limits and Advisor Visibility

Delegation is one level deep and a roster holds at most twenty agents, with twenty-five concurrent threads supported [6] — constraints that shape an architecture rather than merely bound it. And where a session uses an advisor, whether the advice is readable on the event stream is the advisor model’s policy: some advisor models deliver a redaction placeholder to every client surface while the agent reads the full advice server-side [6]. An assurance program that treats the event stream as a complete record of what informed a decision should verify that assumption against the advisor model in use.

Obligation 03  ·  Evaluation, Also Authored

Evaluation is available and is worth naming, because it is authored by the customer too. Outcomes take a rubric describing what success looks like, and a separate grader evaluates the output in its own context window so the assessment is not colored by the agent’s own reasoning [7].

The Hard Claim

Managed Agents is stateful by design, and Anthropic published what follows from that: the product carries no zero-retention or healthcare-agreement eligibility, across every sub-feature, self-hosted sandboxes included. That is the constraint stated in the place it belongs.

The same statefulness is why this platform has the arc’s only currency-denominated ceiling and its only tool gate that pauses a session and returns a reason. A session that persists is a session that can be stopped, asked, resumed, and priced. The architect’s decision is not whether statefulness is acceptable in the abstract. It is which workloads belong on a surface whose record is processed by the platform, and that decision is made at design time against the published statement rather than after an audit asks for it.

Next in the arc: the OpenAI Architect dispatch reads Presence against the same three questions, where each answer is determined per deployment rather than in configuration.

A Published Constraint Is a Design Input, Not a Surprise.

If you are deciding which workloads belong on a stateful agent surface and want a practitioner conversation, the calendar is open.

Start the conversation
The Published Boundary  ·  Series 32  ·  Five Posts
Prologue  ·  Published Every Harness Publishes Its Boundary
Post 01  ·  Published Model Choice Is a Customer Control
Post 02  ·  Now Reading Stateful by Design, and Documented as Such
Post 03  ·  Published Defined for Each Deployment
Post 04  ·  Published Know Who Holds the State
References & Sources
The Published Boundary — Where the Harness Ends  ·  Five Posts  ·  August 2026

Share this:

Like this:

Like Loading…