{
  "fixes": {
    "slug": "gemini-enterprise",
    "name": "Gemini Enterprise",
    "listing": "https://www.anchorterminal.com/tools/gemini-enterprise",
    "markdown": "# Fix list: Gemini Enterprise\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/gemini-enterprise, the October 2026 research run, assessed 10 October 2026. Grade B, 68.1 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 Gemini Enterprise: 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, 10 out of 100, up to 11.3 more on the total\n\nWhy it scored 10: Hosted rubric. No x402, MPP or L402 (0). No seat prices were found on Google's product, editions, licences or quotas pages, and the product page sends buyers to sales or a trial. Pay-as-you-go and overages are priced publicly by reference, $5 per GiB a month for storage over quota and Agent Platform prices for assistant queries, without a per-query figure. Counted as plan-level pricing that's public (10, a judgement call). A 30-day Plus trial needs no sales contact, but the product page sends it through the Google Cloud Free Trial, which asks for a credit card or other payment method (0). A person creates the project, billing account and app in the console and assigns licences (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, 48 out of 100, up to 8.5 more on the total\n\nWhy it scored 48: API reading. StreamAssist streams reply fragments plus planner steps in `diagnosticInfo`, and the request has no setting that limits answer length (10 of 25). `toolsSpec` narrows the call to chosen data stores with a filter expression, `agentsSpec` picks the agent, and session and list methods page with `pageSize` and `pageToken` (14 of 20). Errors follow Google's status and message model, and the response lists `connectorAuthErrors` per connector so a client can see which source was skipped. No Gemini Enterprise error page was found (12 of 20). No idempotency key, and no safe-retry guidance was found. `assistToken` identifies a call for feedback and support only (0 of 20). Only `query` is needed once an app exists, but the app, a licence and the engine ID come first. Client libraries for C#, Go, Java, Node.js, PHP, Python and Ruby (12 of 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, 78 out of 100, up to 3.6 more on the total\n\nWhy it scored 78: A public Discovery document for v1, revision 20261004, with 278 methods and 990 schemas, StreamAssist and the A2A methods among them (25). docs.cloud.google.com/llms.txt answers 404. A page address with `.md.txt` added answers as Markdown, which no page we read links, so half (5 of 10). StreamAssist's description is one line, while the guide says when to use the API rather than the web app and what sessions and files do (12 of 20). 3,421 of 3,448 properties carry a description and 245 are enums. Required and optional are stated in each description, and `inputVariables` for workflow agents is a free Struct (11 of 15). The guide has a curl request with a full streamed response and the reference covers seven client libraries. No Gemini Enterprise page listing error codes was found (10 of 15). v1, v1beta and v1alpha versions and dated release notes, the newest on 8 October 2026 (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. Reliability, 85 out of 100, up to 3 more on the total\n\nWhy it scored 85: Hosted reading, scored on the StreamAssist method of the Discovery Engine API. The Google Cloud status page lists Gemini Enterprise as a product and has a JSON incident feed (20). The feed's five incidents since 10 July 2026 (the latest on 8 October, Cloud Storage) name no Gemini Enterprise, Agent Search or Vertex AI Search product, and the page shows only broad incidents (30). Rate quotas with numbers, 300 search and complete-query requests a minute per project, and assistant quotas per licence, 160 queries a day on Standard and 200 on Plus, pooled per project and location (15). No retry or backoff guidance and no 429 handling were found on the StreamAssist, quotas or A2A pages, and each call writes to a session, so a retry isn't documented as safe (0 of 15). The Gemini Enterprise SLA commits to 99.5 per cent monthly uptime for StreamAssist and 99.9 for Search, with credits from 10 to 50 per cent (10). StreamAssist is in the v1 API and its guide marks only file upload as Preview (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## 5. Security \u0026 auth, 87 out of 100, up to 2.3 more on the total\n\nWhy it scored 87: OAuth 2.0 bearer tokens from user credentials, service accounts or workforce identity federation, with IAM permissions. StreamAssist accepts the narrow `discoveryengine.assist.readwrite` scope among four, and the docs send the token only in the `Authorization` header (30). Gemini Enterprise User can be granted per app through the Restricted User pattern, a custom role can hold only `discoveryengine.assistants.assist`, and VPC Service Controls apply. The assistant can run connector actions such as sending Outlook mail, and we didn't read how it asks for confirmation (14 of 20). Model Armor screens the assistant's prompts and responses on every edition at no extra cost, but it's off until an admin enables it and can be set to let traffic through unscreened when it's unavailable (10 of 15). Cloud Audit Logs list each method by permission type, and usage audit logs can record prompts and responses when that setting is on. Data Access logs are off by default in Google Cloud (13 of 15). google.com security.txt is valid to 1 April 2030 with the reward programme, and the compliance page lists ISO 27001, 27017, 27018 and 27701, SOC 1, 2 and 3, HIPAA, FedRAMP, PCI DSS and C5 for the Standard and Plus editions (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## 6. Maintenance \u0026 community, 80 out of 100, up to 1.8 more on the total\n\nWhy it scored 80: Read as a closed service with official client libraries. Newest release notes dated 8 October 2026 (Claude models for AI developer tools, Gemini 3.8 Flash in Singapore) (30). Dozens of dated release note entries in September and October 2026 alone (20). Dated release notes, Google Cloud support and public issue trackers for the client libraries. We didn't test support (10 of 15). Official client libraries, PyPI `google-cloud-discoveryengine` 0.20.5 on 1 October 2026 and npm `@google-cloud/discoveryengine` 3.3.1 (15). The Python library is still below 1.0 and declares Python 3.10 or later. CI results weren't checked (5 of 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## 7. Transparency \u0026 trust, 82 out of 100, up to 1.6 more on the total\n\nMade of editorial 64, provenance 99.\n\nWhy it scored 82: Closed service under the Google Cloud terms, with Apache-2.0 client libraries (15 of 30). The product page says prompts, outputs and tuning are never used to train Google models or other customers' models, and the security overview says user-requested data is deleted within 60 days. The Cloud terms and Data Processing Addendum weren't re-read today, and retention for sessions and memory wasn't found (20 of 30). The Cloud terms promise 12 months' notice before a backwards-incompatible API change, per the 9 October Document AI check, not re-read on 10 October. The release notes gave 19 days' notice (1 to 20 October) for removing Gemini 2.5 Pro from the Canada and Japan regions (12 of 20). Global, `us` and `eu` multi-regions with data residency limits stated, and Google Cloud's public sub-processor list per the 9 October check (17 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- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)\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 Gemini agent announced on 8 October 2026 (coworker agents with their own Workspace accounts, sub-agents, routing to Claude) is reachable through StreamAssist or another API. The announcement says 'a single API' and headless operation, but no API page for it was found on 10 October\n- Seat prices for each edition. None of the Google pages read gives a figure\n- How the assistant confirms connector write actions, such as sending Outlook mail, when called through StreamAssist\n- Retention of sessions, memory and uploaded files beyond the 60-day deletion of user-requested data\n- unchecked: the Google Cloud terms' 12-month change notice, the Data Processing Addendum and the sub-processor list, kept from the 9 October Document AI check\n- unchecked: GitHub issue activity and CI for the Discovery Engine client libraries\n- Disclosure. These grades were written by agents running on Anthropic's Claude, and Gemini Enterprise routes some work to Claude models. The judgement calls are 10 for pricing that's public only by reference to Agent Platform rates, and 0 for retry handling because none was found\n\n## Weaknesses\n\n- No seat prices on the Google pages read, and pay-as-you-go needs an invoiced Cloud Billing account\n- The 30-day Plus trial goes through the Google Cloud Free Trial, which asks for a card or other payment method\n- No retry, backoff or idempotency guidance for StreamAssist was found, and each call writes to a session\n- Background tasks, desktop and mobile apps and third-party model choice are early access per the product page\n- Assistant use is capped per licence, 160 queries a day on Standard and 200 on Plus, pooled per project\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- Create a Gemini Enterprise app first. StreamAssist needs its APP_ID, and the caller needs `discoveryengine.assistants.assist`\n- Send `X-Goog-User-Project` with your project ID when calling with Application Default Credentials\n- Pass the `session` name from the final stream message on the next call to keep context. Leaving it empty starts a new session\n- Read `connectorAuthErrors` in the response. A connector that failed sign-in is skipped and the answer comes from the rest\n- Use the A2A URL returned by Agent Registry verbatim. It takes the project number, not the project ID\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": "B",
    "score": 68.1,
    "assessed": "2026-10-10",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 10,
        "maxGain": 11.3,
        "reason": "Hosted rubric. No x402, MPP or L402 (0). No seat prices were found on Google's product, editions, licences or quotas pages, and the product page sends buyers to sales or a trial. Pay-as-you-go and overages are priced publicly by reference, $5 per GiB a month for storage over quota and Agent Platform prices for assistant queries, without a per-query figure. Counted as plan-level pricing that's public (10, a judgement call). A 30-day Plus trial needs no sales contact, but the product page sends it through the Google Cloud Free Trial, which asks for a credit card or other payment method (0). A person creates the project, billing account and app in the console and assigns licences (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": 48,
        "maxGain": 8.5,
        "reason": "API reading. StreamAssist streams reply fragments plus planner steps in `diagnosticInfo`, and the request has no setting that limits answer length (10 of 25). `toolsSpec` narrows the call to chosen data stores with a filter expression, `agentsSpec` picks the agent, and session and list methods page with `pageSize` and `pageToken` (14 of 20). Errors follow Google's status and message model, and the response lists `connectorAuthErrors` per connector so a client can see which source was skipped. No Gemini Enterprise error page was found (12 of 20). No idempotency key, and no safe-retry guidance was found. `assistToken` identifies a call for feedback and support only (0 of 20). Only `query` is needed once an app exists, but the app, a licence and the engine ID come first. Client libraries for C#, Go, Java, Node.js, PHP, Python and Ruby (12 of 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": 78,
        "maxGain": 3.6,
        "reason": "A public Discovery document for v1, revision 20261004, with 278 methods and 990 schemas, StreamAssist and the A2A methods among them (25). docs.cloud.google.com/llms.txt answers 404. A page address with `.md.txt` added answers as Markdown, which no page we read links, so half (5 of 10). StreamAssist's description is one line, while the guide says when to use the API rather than the web app and what sessions and files do (12 of 20). 3,421 of 3,448 properties carry a description and 245 are enums. Required and optional are stated in each description, and `inputVariables` for workflow agents is a free Struct (11 of 15). The guide has a curl request with a full streamed response and the reference covers seven client libraries. No Gemini Enterprise page listing error codes was found (10 of 15). v1, v1beta and v1alpha versions and dated release notes, the newest on 8 October 2026 (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": "reliability",
        "name": "Reliability",
        "score": 85,
        "maxGain": 3,
        "reason": "Hosted reading, scored on the StreamAssist method of the Discovery Engine API. The Google Cloud status page lists Gemini Enterprise as a product and has a JSON incident feed (20). The feed's five incidents since 10 July 2026 (the latest on 8 October, Cloud Storage) name no Gemini Enterprise, Agent Search or Vertex AI Search product, and the page shows only broad incidents (30). Rate quotas with numbers, 300 search and complete-query requests a minute per project, and assistant quotas per licence, 160 queries a day on Standard and 200 on Plus, pooled per project and location (15). No retry or backoff guidance and no 429 handling were found on the StreamAssist, quotas or A2A pages, and each call writes to a session, so a retry isn't documented as safe (0 of 15). The Gemini Enterprise SLA commits to 99.5 per cent monthly uptime for StreamAssist and 99.9 for Search, with credits from 10 to 50 per cent (10). StreamAssist is in the v1 API and its guide marks only file upload as Preview (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": "security",
        "name": "Security \u0026 auth",
        "score": 87,
        "maxGain": 2.3,
        "reason": "OAuth 2.0 bearer tokens from user credentials, service accounts or workforce identity federation, with IAM permissions. StreamAssist accepts the narrow `discoveryengine.assist.readwrite` scope among four, and the docs send the token only in the `Authorization` header (30). Gemini Enterprise User can be granted per app through the Restricted User pattern, a custom role can hold only `discoveryengine.assistants.assist`, and VPC Service Controls apply. The assistant can run connector actions such as sending Outlook mail, and we didn't read how it asks for confirmation (14 of 20). Model Armor screens the assistant's prompts and responses on every edition at no extra cost, but it's off until an admin enables it and can be set to let traffic through unscreened when it's unavailable (10 of 15). Cloud Audit Logs list each method by permission type, and usage audit logs can record prompts and responses when that setting is on. Data Access logs are off by default in Google Cloud (13 of 15). google.com security.txt is valid to 1 April 2030 with the reward programme, and the compliance page lists ISO 27001, 27017, 27018 and 27701, SOC 1, 2 and 3, HIPAA, FedRAMP, PCI DSS and C5 for the Standard and Plus editions (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"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 80,
        "maxGain": 1.8,
        "reason": "Read as a closed service with official client libraries. Newest release notes dated 8 October 2026 (Claude models for AI developer tools, Gemini 3.8 Flash in Singapore) (30). Dozens of dated release note entries in September and October 2026 alone (20). Dated release notes, Google Cloud support and public issue trackers for the client libraries. We didn't test support (10 of 15). Official client libraries, PyPI `google-cloud-discoveryengine` 0.20.5 on 1 October 2026 and npm `@google-cloud/discoveryengine` 3.3.1 (15). The Python library is still below 1.0 and declares Python 3.10 or later. CI results weren't checked (5 of 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": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 82,
        "maxGain": 1.6,
        "reason": "Closed service under the Google Cloud terms, with Apache-2.0 client libraries (15 of 30). The product page says prompts, outputs and tuning are never used to train Google models or other customers' models, and the security overview says user-requested data is deleted within 60 days. The Cloud terms and Data Processing Addendum weren't re-read today, and retention for sessions and memory wasn't found (20 of 30). The Cloud terms promise 12 months' notice before a backwards-incompatible API change, per the 9 October Document AI check, not re-read on 10 October. The release notes gave 19 days' notice (1 to 20 October) for removing Gemini 2.5 Pro from the Canada and Japan regions (12 of 20). Global, `us` and `eu` multi-regions with data residency limits stated, and Google Cloud's public sub-processor list per the 9 October check (17 of 20).",
        "blend": "editorial 64, provenance 99",
        "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"
      }
    ],
    "provenance": [
      {
        "label": "Privacy policy",
        "value": "read, states 7 of the 8 things a reader expects",
        "points": 9.3,
        "max": 10
      }
    ],
    "unchecked": [
      "Whether the Gemini agent announced on 8 October 2026 (coworker agents with their own Workspace accounts, sub-agents, routing to Claude) is reachable through StreamAssist or another API. The announcement says 'a single API' and headless operation, but no API page for it was found on 10 October",
      "Seat prices for each edition. None of the Google pages read gives a figure",
      "How the assistant confirms connector write actions, such as sending Outlook mail, when called through StreamAssist",
      "Retention of sessions, memory and uploaded files beyond the 60-day deletion of user-requested data",
      "unchecked: the Google Cloud terms' 12-month change notice, the Data Processing Addendum and the sub-processor list, kept from the 9 October Document AI check",
      "unchecked: GitHub issue activity and CI for the Discovery Engine client libraries",
      "Disclosure. These grades were written by agents running on Anthropic's Claude, and Gemini Enterprise routes some work to Claude models. The judgement calls are 10 for pricing that's public only by reference to Agent Platform rates, and 0 for retry handling because none was found"
    ],
    "weaknesses": [
      "No seat prices on the Google pages read, and pay-as-you-go needs an invoiced Cloud Billing account",
      "The 30-day Plus trial goes through the Google Cloud Free Trial, which asks for a card or other payment method",
      "No retry, backoff or idempotency guidance for StreamAssist was found, and each call writes to a session",
      "Background tasks, desktop and mobile apps and third-party model choice are early access per the product page",
      "Assistant use is capped per licence, 160 queries a day on Standard and 200 on Plus, pooled per project"
    ],
    "agentNotes": [
      "Create a Gemini Enterprise app first. StreamAssist needs its APP_ID, and the caller needs `discoveryengine.assistants.assist`",
      "Send `X-Goog-User-Project` with your project ID when calling with Application Default Credentials",
      "Pass the `session` name from the final stream message on the next call to keep context. Leaving it empty starts a new session",
      "Read `connectorAuthErrors` in the response. A connector that failed sign-in is skipped and the answer comes from the rest",
      "Use the A2A URL returned by Agent Registry verbatim. It takes the project number, not the project ID"
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-11",
    "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"
  }
}