{
  "fixes": {
    "slug": "agentcore-code-interpreter",
    "name": "Amazon Bedrock AgentCore Code Interpreter",
    "listing": "https://www.anchorterminal.com/tools/agentcore-code-interpreter",
    "markdown": "# Fix list: Amazon Bedrock AgentCore Code Interpreter\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/agentcore-code-interpreter, the October 2026 research run, assessed 8 October 2026. Grade BB, 73.1 out of 100.\n\nThis is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.\n\nFor a coding agent working on Amazon Bedrock AgentCore Code Interpreter: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.\n\n## 1. Payments \u0026 pricing, 30 out of 100, up to 8.8 more on the total\n\nWhy it scored 30: No x402, MPP or L402 on the Code Interpreter endpoints. AgentCore Payments is a separate component for an agent paying third parties (0). Per-unit prices published without a login, $0.0895 a vCPU-hour and $0.00945 a GB-hour (20). No allowance for Code Interpreter, but the pricing page says new AWS customers get up to $200 in Free Tier credits and the Free Tier FAQ says most new customers need no payment method, so half, as on the other AWS listings (10). A person signs up in a browser and sets up IAM (0).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):\n\nThe published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).\n\n- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).\n\nPayment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.\n\nOpen-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.\n\n## 2. Agent ergonomics, 75 out of 100, up to 4.1 more on the total\n\nWhy it scored 75: Results stream back as content blocks with stdout and stderr, and we found no control for truncating or sizing output. The MCP server registers 122 tools by default and `AGENTCORE_ENABLE_TOOLS` cuts that to the 10 for Code Interpreter (15 of 25). `ListCodeInterpreterSessions` pages with `maxResults` of 1 to 100 and `nextToken` and filters by status (15 of 20). Typed errors with HTTP codes, but a quota breach comes back as HTTP 402 and errors can also arrive inside the result stream (15 of 20). `clientToken` on session start and on interpreter create and delete. Running code has no idempotency key (15 of 20). The managed interpreter needs no set-up, `code_session` handles start and stop, and there are AgentCore SDKs for Python and TypeScript beside the AWS SDKs (15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):\n\n- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.\n\nModels are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.\n\n## 3. Schema \u0026 documentation, 81 out of 100, up to 3.1 more on the total\n\nWhy it scored 81: The API model is public as Smithy JSON in aws/api-models-aws, version 2024-02-28, and the SDKs are generated from it (25). The developer guide has an llms.txt of 988 lines and serves every page as Markdown (10). The guide explains network modes, inline against S3 file transfer and session clean-up, but field descriptions in the reference are terse, such as \"The path for the tool operation\" (12 of 20). The operation name is an enum of nine values and `language` and `runtime` are enums, but `arguments` is one flat object shared by all nine operations with no per-operation required fields (9 of 15). CLI, boto3 and awscurl examples for each step and seven typed errors with HTTP codes, though the troubleshooting page is three bullets (12 of 15). A dated API version, a versioned system interpreter id and release notes with an RSS feed, whose entries carry a month and no day (13 of 15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):\n\nAPIs and MCP servers.\n\n- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.\n\nModels are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.\n\n## 4. Maintenance \u0026 community, 65 out of 100, up to 3.1 more on the total\n\nWhy it scored 65: The Python SDK that carries `code_session` shipped 1.24.1 on 7 October 2026, but the newest changes to Code Interpreter itself are from July 2026, the `ActiveSessionCount` metric and a fix to `install_packages` on 17 July, so 20 of 30 as a judgement call (20). The SDK had ten releases since 10 July and the release notes have entries every month, yet only two touch Code Interpreter (10 of 20). Public release notes with RSS, issues open on the SDK repository with an auto-triage workflow, and re:Post and AWS Support. Reply times were not checked (10 of 15). Current official SDKs, `bedrock-agentcore` 1.24.1 on PyPI and 0.4.5 on npm (15). The SDK repository runs CI, integration tests, security scanning and a breaking-change check, with a lock file (10).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):\n\n- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.\n\nModels are read for deprecation notice periods and model churn rather than release counts.\n\n## 5. Reliability, 85 out of 100, up to 3 more on the total\n\nWhy it scored 85: Graded on the hosted lines, for the AWS API. AWS Health Dashboard with per-service, per-Region history (20). The dashboard history file, read on 8 October, names AgentCore in one event in the 90 days, elevated packet loss in one availability zone of Europe (Spain) for 2 hours 40 minutes on 4 October 2026, marked informational. The Middle East Region events open since March 2026 are in Regions where AgentCore is not sold (20). Rate limits published per API, 30 requests a second for starting sessions and invoking and 5 for creating or deleting interpreters, with 1,000 concurrent sessions (15). `ThrottlingException` is a 429 with advice to back off exponentially, `InternalServerException` is documented as retryable, and `StartCodeInterpreterSession` takes a `clientToken` so a retry does not open a second session (15). The AgentCore FAQ says the Bedrock SLA applies, but that SLA, last updated 4 October 2023, names only the Bedrock APIs for models (5 of 10, our call, as on the Bedrock Guardrails listing). Generally available since October 2025 (10).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):\n\nHosted APIs, MCP servers, models and platforms.\n\n- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.\n\nLocal packages, SDKs, frameworks and stdio MCP servers.\n\n- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.\n\nProtocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.\n\n## 6. Transparency \u0026 trust, 70 out of 100, up to 2.6 more on the total\n\nMade of editorial 51, provenance 88.\n\nWhy it scored 70: Closed service under the AWS Service Terms, with Apache-2.0 SDKs and MCP server (15). The docs say session files are cleaned up when a session ends and also give a 30-day retention period for session data without saying what is kept. Encryption at rest and TLS 1.2 are stated, the privacy notice excludes customer content, and we found no Code Interpreter statement on use of customer content (15 of 30). No deprecation policy for the interpreter or its language runtimes was found. The system interpreter id carries a version (3 of 20). AWS publishes a sub-processor list, updated 28 July 2026, and the guide lists the 22 Regions where built-in tools run (18 of 20).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):\n\n- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).\n\nThe other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.\n\nProvenance checks not met in full (half of this category, computed from checked facts):\n\n- Terms of service: read, states 6 of the 7 things a reader expects, and has 3 clauses that cost points (3.1 of 10)\n- security.txt: published but past its Expires date (5 of 10)\n\n## 7. Security \u0026 auth, 87 out of 100, up to 2.3 more on the total\n\nWhy it scored 87: IAM with SigV4, roles and short-lived credentials, and policies can name one interpreter ARN (30). Each of the nine operations is its own IAM action, system and custom interpreters have separate ARN types, a custom interpreter's execution role bounds what sandbox code can reach, and the default Sandbox network mode reaches only Amazon S3. No approval step for destructive calls (17 of 20). Each session is a dedicated microVM whose memory AWS says is sanitised at the end, and the docs flag Public network mode as a risk, but the Code Interpreter pages we read give no guidance on treating execution output as untrusted (11 of 15). CloudWatch metrics, spans and per-second usage logs per session, and the overview claims CloudTrail logging, but we found no page listing Code Interpreter's CloudTrail events, and console logs are not sent to CloudWatch (11 of 15). A vulnerability disclosure programme on HackerOne, public bulletins and AgentCore in SOC scope on the list of 11 August 2026. The security.txt expired on 24 September 2026 and no paid bounty was found, read as on the other AWS listings (18 of 20).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-security):\n\n- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.\n\nModels are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.\n\n## What we couldn't check\n\nWhat we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.\n\n- Whether the Bedrock SLA's wording, the Bedrock APIs for models, covers Code Interpreter. The AgentCore FAQ says the SLA applies and the SLA text does not name AgentCore.\n- What the 30-day retention period for session data covers, since the same page says session data is cleaned up when the session ends.\n- Which Code Interpreter calls are recorded in CloudTrail, and whether `InvokeCodeInterpreter` is a data event. The overview claims CloudTrail logging and we found no event list.\n- The day in July 2026 on which the `ActiveSessionCount` metric shipped. The release notes carry months only.\n- `firstReleased` is the date of the first `bedrock-agentcore` release on PyPI, 8 July 2025. The preview announcement itself was not read.\n- unchecked: weekly PyPI downloads for `bedrock-agentcore`. pypistats.org answered 429 on the first request and we did not retry.\n- unchecked: the AWS Data Processing Addendum and the AWS Customer Agreement were not read this run.\n- unchecked: the MCP server's tool definitions and annotations. Only its README was read.\n- unchecked: reply times on the SDK repositories' issues. The GitHub API gave 776 stars and 135 open issues and pull requests for the Python SDK.\n\n## Weaknesses\n\n- No pause, resume or snapshot. Session files are removed when the session ends, and persistence needs a customer-owned S3 Files or EFS mount inside a VPC\n- Every session is capped at 2 vCPU, 8 GB of memory and 10 GB of disk, and the cap is not adjustable\n- `InvokeCodeInterpreter` takes one flat `arguments` object for nine operations, so the reference does not say which fields each operation requires\n- The Bedrock SLA that AWS says applies to AgentCore dates from 4 October 2023 and names only the Bedrock APIs for models\n- The troubleshooting page is three bullet points, and execution console logs are not sent to CloudWatch\n\n## What costs an agent a turn today\n\nThe notes we give agents before they call it. Each one is a workaround an agent shouldn't need.\n\n- Start a session with `StartCodeInterpreterSession`, then pass its id in the `x-amzn-code-interpreter-session-id` header on every `InvokeCodeInterpreter` call\n- Set `sessionTimeoutSeconds` when starting. The default is 900 seconds and the maximum is eight hours, and the session ends itself at the timeout\n- Stop sessions when done. Billing runs per second while code is busy, and a session left open counts against the 1,000 concurrent-session quota\n- Use `startCommandExecution`, `getTask` and `stopTask` for work longer than the 15-minute synchronous request limit\n- Retry `ThrottlingException` (429) and `InternalServerException` (500) with exponential backoff, and treat `ServiceQuotaExceededException`, returned as HTTP 402, as a quota to raise\n\n## When it's done\n\nSend what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `\"kind\": \"dispute\"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.\n",
    "grade": "BB",
    "score": 73.1,
    "assessed": "2026-10-08",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 30,
        "maxGain": 8.8,
        "reason": "No x402, MPP or L402 on the Code Interpreter endpoints. AgentCore Payments is a separate component for an agent paying third parties (0). Per-unit prices published without a login, $0.0895 a vCPU-hour and $0.00945 a GB-hour (20). No allowance for Code Interpreter, but the pricing page says new AWS customers get up to $200 in Free Tier credits and the Free Tier FAQ says most new customers need no payment method, so half, as on the other AWS listings (10). A person signs up in a browser and sets up IAM (0).",
        "checklist": [
          "The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).",
          "- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).",
          "Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.",
          "Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-payments"
      },
      {
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "score": 75,
        "maxGain": 4.1,
        "reason": "Results stream back as content blocks with stdout and stderr, and we found no control for truncating or sizing output. The MCP server registers 122 tools by default and `AGENTCORE_ENABLE_TOOLS` cuts that to the 10 for Code Interpreter (15 of 25). `ListCodeInterpreterSessions` pages with `maxResults` of 1 to 100 and `nextToken` and filters by status (15 of 20). Typed errors with HTTP codes, but a quota breach comes back as HTTP 402 and errors can also arrive inside the result stream (15 of 20). `clientToken` on session start and on interpreter create and delete. Running code has no idempotency key (15 of 20). The managed interpreter needs no set-up, `code_session` handles start and stop, and there are AgentCore SDKs for Python and TypeScript beside the AWS SDKs (15).",
        "checklist": [
          "- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.",
          "Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-ergonomics"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 81,
        "maxGain": 3.1,
        "reason": "The API model is public as Smithy JSON in aws/api-models-aws, version 2024-02-28, and the SDKs are generated from it (25). The developer guide has an llms.txt of 988 lines and serves every page as Markdown (10). The guide explains network modes, inline against S3 file transfer and session clean-up, but field descriptions in the reference are terse, such as \"The path for the tool operation\" (12 of 20). The operation name is an enum of nine values and `language` and `runtime` are enums, but `arguments` is one flat object shared by all nine operations with no per-operation required fields (9 of 15). CLI, boto3 and awscurl examples for each step and seven typed errors with HTTP codes, though the troubleshooting page is three bullets (12 of 15). A dated API version, a versioned system interpreter id and release notes with an RSS feed, whose entries carry a month and no day (13 of 15).",
        "checklist": [
          "APIs and MCP servers.",
          "- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.",
          "Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-schema"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 65,
        "maxGain": 3.1,
        "reason": "The Python SDK that carries `code_session` shipped 1.24.1 on 7 October 2026, but the newest changes to Code Interpreter itself are from July 2026, the `ActiveSessionCount` metric and a fix to `install_packages` on 17 July, so 20 of 30 as a judgement call (20). The SDK had ten releases since 10 July and the release notes have entries every month, yet only two touch Code Interpreter (10 of 20). Public release notes with RSS, issues open on the SDK repository with an auto-triage workflow, and re:Post and AWS Support. Reply times were not checked (10 of 15). Current official SDKs, `bedrock-agentcore` 1.24.1 on PyPI and 0.4.5 on npm (15). The SDK repository runs CI, integration tests, security scanning and a breaking-change check, with a lock file (10).",
        "checklist": [
          "- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.",
          "Models are read for deprecation notice periods and model churn rather than release counts."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-maintenance"
      },
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 85,
        "maxGain": 3,
        "reason": "Graded on the hosted lines, for the AWS API. AWS Health Dashboard with per-service, per-Region history (20). The dashboard history file, read on 8 October, names AgentCore in one event in the 90 days, elevated packet loss in one availability zone of Europe (Spain) for 2 hours 40 minutes on 4 October 2026, marked informational. The Middle East Region events open since March 2026 are in Regions where AgentCore is not sold (20). Rate limits published per API, 30 requests a second for starting sessions and invoking and 5 for creating or deleting interpreters, with 1,000 concurrent sessions (15). `ThrottlingException` is a 429 with advice to back off exponentially, `InternalServerException` is documented as retryable, and `StartCodeInterpreterSession` takes a `clientToken` so a retry does not open a second session (15). The AgentCore FAQ says the Bedrock SLA applies, but that SLA, last updated 4 October 2023, names only the Bedrock APIs for models (5 of 10, our call, as on the Bedrock Guardrails listing). Generally available since October 2025 (10).",
        "checklist": [
          "Hosted APIs, MCP servers, models and platforms.",
          "- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.",
          "Local packages, SDKs, frameworks and stdio MCP servers.",
          "- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.",
          "Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-reliability"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 70,
        "maxGain": 2.6,
        "reason": "Closed service under the AWS Service Terms, with Apache-2.0 SDKs and MCP server (15). The docs say session files are cleaned up when a session ends and also give a 30-day retention period for session data without saying what is kept. Encryption at rest and TLS 1.2 are stated, the privacy notice excludes customer content, and we found no Code Interpreter statement on use of customer content (15 of 30). No deprecation policy for the interpreter or its language runtimes was found. The system interpreter id carries a version (3 of 20). AWS publishes a sub-processor list, updated 28 July 2026, and the guide lists the 22 Regions where built-in tools run (18 of 20).",
        "blend": "editorial 51, provenance 88",
        "checklist": [
          "- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).",
          "The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-transparency"
      },
      {
        "key": "security",
        "name": "Security \u0026 auth",
        "score": 87,
        "maxGain": 2.3,
        "reason": "IAM with SigV4, roles and short-lived credentials, and policies can name one interpreter ARN (30). Each of the nine operations is its own IAM action, system and custom interpreters have separate ARN types, a custom interpreter's execution role bounds what sandbox code can reach, and the default Sandbox network mode reaches only Amazon S3. No approval step for destructive calls (17 of 20). Each session is a dedicated microVM whose memory AWS says is sanitised at the end, and the docs flag Public network mode as a risk, but the Code Interpreter pages we read give no guidance on treating execution output as untrusted (11 of 15). CloudWatch metrics, spans and per-second usage logs per session, and the overview claims CloudTrail logging, but we found no page listing Code Interpreter's CloudTrail events, and console logs are not sent to CloudWatch (11 of 15). A vulnerability disclosure programme on HackerOne, public bulletins and AgentCore in SOC scope on the list of 11 August 2026. The security.txt expired on 24 September 2026 and no paid bounty was found, read as on the other AWS listings (18 of 20).",
        "checklist": [
          "- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.",
          "Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-security"
      }
    ],
    "provenance": [
      {
        "label": "Terms of service",
        "value": "read, states 6 of the 7 things a reader expects, and has 3 clauses that cost points",
        "points": 3.1,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "published but past its Expires date",
        "points": 5,
        "max": 10
      }
    ],
    "unchecked": [
      "Whether the Bedrock SLA's wording, the Bedrock APIs for models, covers Code Interpreter. The AgentCore FAQ says the SLA applies and the SLA text does not name AgentCore.",
      "What the 30-day retention period for session data covers, since the same page says session data is cleaned up when the session ends.",
      "Which Code Interpreter calls are recorded in CloudTrail, and whether `InvokeCodeInterpreter` is a data event. The overview claims CloudTrail logging and we found no event list.",
      "The day in July 2026 on which the `ActiveSessionCount` metric shipped. The release notes carry months only.",
      "`firstReleased` is the date of the first `bedrock-agentcore` release on PyPI, 8 July 2025. The preview announcement itself was not read.",
      "unchecked: weekly PyPI downloads for `bedrock-agentcore`. pypistats.org answered 429 on the first request and we did not retry.",
      "unchecked: the AWS Data Processing Addendum and the AWS Customer Agreement were not read this run.",
      "unchecked: the MCP server's tool definitions and annotations. Only its README was read.",
      "unchecked: reply times on the SDK repositories' issues. The GitHub API gave 776 stars and 135 open issues and pull requests for the Python SDK."
    ],
    "weaknesses": [
      "No pause, resume or snapshot. Session files are removed when the session ends, and persistence needs a customer-owned S3 Files or EFS mount inside a VPC",
      "Every session is capped at 2 vCPU, 8 GB of memory and 10 GB of disk, and the cap is not adjustable",
      "`InvokeCodeInterpreter` takes one flat `arguments` object for nine operations, so the reference does not say which fields each operation requires",
      "The Bedrock SLA that AWS says applies to AgentCore dates from 4 October 2023 and names only the Bedrock APIs for models",
      "The troubleshooting page is three bullet points, and execution console logs are not sent to CloudWatch"
    ],
    "agentNotes": [
      "Start a session with `StartCodeInterpreterSession`, then pass its id in the `x-amzn-code-interpreter-session-id` header on every `InvokeCodeInterpreter` call",
      "Set `sessionTimeoutSeconds` when starting. The default is 900 seconds and the maximum is eight hours, and the session ends itself at the timeout",
      "Stop sessions when done. Billing runs per second while code is busy, and a session left open counts against the 1,000 concurrent-session quota",
      "Use `startCommandExecution`, `getTask` and `stopTask` for work longer than the 15-minute synchronous request limit",
      "Retry `ThrottlingException` (429) and `InternalServerException` (500) with exponential backoff, and treat `ServiceQuotaExceededException`, returned as HTTP 402, as a quota to raise"
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-09",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.4",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}