Dots

# Fix list: Dots

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/openai-dots, the October 2026 research run, assessed 12 October 2026. Grade F, 28.3 out of 100.

This 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.

For a coding agent working on Dots: 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.

## 1. Reliability, 3 out of 100, up to 19.4 more on the total

Why it scored 3: Read as a hosted service. No status page is linked from the Dots docs, the admin guide, the pricing page or the release notes, and the hosts where one might be linked refused our reader, so both status lines score as absent and are listed as unchecked (0 and 0). The access section mentions an allowance for deeper work and extended limits for the first month, with no numbers (0). The docs say Pause stops only the main task, that stopping does not undo completed actions and that a completed run does not confirm the result was achieved or delivered. No retry or overload guidance was found (3 of 15). No SLA found (0). The admin setting is labelled Use dots (Beta), access is rolling out gradually and Teams is an invite-only alpha (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 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.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 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.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 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.

## 2. Agent ergonomics, 22 out of 100, up to 12.7 more on the total

Why it scored 22: No programmatic interface, so an outside agent cannot call a dot and these lines are read for the person directing it. The owner can say which changes deserve a notification and where updates go, with no stated control over answer size (5 of 25). Activity and Scheduled views list delegated and recurring tasks in the desktop app (3 of 20). Requests for a decision, app connection, sign-in or approval appear in Activity, and no error list is documented (5 of 20). The docs warn that stopping does not undo completed actions and give no retry or duplicate-action guidance (3 of 20). Setup can keep the defaults and skip plugins, and there is no SDK (6 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 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).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 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.

## 3. Payments & pricing, 10 out of 100, up to 11.3 more on the total

Why it scored 10: No x402, MPP or L402 (0). Plan prices are public on the docs pricing page, Pro at $100, $200 or $500 a month, with no figure for the dots allowance and no price found for Business Premium (10). Dots is not listed for the Free, Go or Plus plans and no trial was found (0). A person signs in to ChatGPT and creates the dot in an app or browser (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

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.
- 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.
- 20, a free tier or trial that doesn't need a card.
- 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.

## 4. Schema & documentation, 41 out of 100, up to 9.6 more on the total

Why it scored 41: Dots has no API or tool definitions, so this is graded on the documentation a person or a model would work from. No machine-readable contract (0). learn.chatgpt.com/llms.txt lists the six Dots pages and the admin guide, each with a Markdown twin (10). The overview says what a dot is for, compares it with Chat, ChatGPT Work and Codex by when to use each, and has a table of what each kind of connection enables (14 of 20). The only enumerated inputs are the four custom rule settings (2 of 15). Example prompts for a first responsibility, a schedule and event monitoring, with no documented errors or failure states (6 of 15). A weekly What's new page with dated entries, three of them on dots, and no version numbers (9 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 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.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 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.

## 5. Security & auth, 58 out of 100, up to 7.4 more on the total

Why it scored 58: Access is through the owner's ChatGPT sign-in, which we could not read about. App connections run under each plugin's connected account and allowed actions, website passwords go to the cloud browser through a private form outside the conversation, reuse of a saved login needs confirmation, and local computer access is revocable. Where tokens and saved passwords are stored is not stated (20 of 30). An automatic review precedes any action that could affect accounts or share information, custom rules have four settings including hand-off, app permissions can allow reading mail without sending, and proactive research tools cannot send or change anything. The docs say rules are instructions the dot tries to follow and can get wrong (16 of 20). The admin guide says built-in safeguards defend against malicious instructions and can pause or stop work, and that messages from other people in Slack do not start work. No further guidance was found (9 of 15). An Activity view for the owner, and Compliance API records of user messages and replies for Enterprise, with coverage the docs tell admins to confirm (9 of 15). A security whitepaper for ChatGPT Work is published and does not mention dots. security.txt answers 404 on the docs host, and no disclosure policy, bounty or certification was on a page we could read (4 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 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.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 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.
- 0 to 15, audit logs or per-call visibility for the operator.
- 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.

## 6. Transparency & trust, 31 out of 100, up to 6 more on the total

Made of editorial 26, provenance 36.

Why it scored 31: Closed service. No page we read names the terms that govern Dots, and openai.com answered 403, so the terms line scores as absent and is listed as unchecked (0 of 30). The docs say ChatGPT data controls apply, that OpenAI does not train directly on proactive research or private notes, and that Enterprise data is not used to improve models by default. The ChatGPT Work cloud security page gives deletion within 30 days for deleted chats and a separate lifecycle for execution state. For a dot's own notes no retention period or per-note deletion is given, and the admin guide's reset link lands on the delete section. The privacy policy was not read (12 of 30). A feature maturity page defines Beta and says anything marked Stable is typically removed through a deprecation process, and the pricing page gives a dated model retirement. No policy for dots itself (8 of 20). Work runs on OpenAI-managed infrastructure with OpenAI's own model named. No sub-processor list or hosting location was on a page we could read (6 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 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.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 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.

Provenance checks not met in full (half of this category, computed from checked facts):

- Legal entity named: not found (0 of 20)
- Domain age: chatgpt.com, registered 2022-11-30 (3 years) (7 of 15)
- Status page: not found (0 of 10)
- security.txt: could not be fetched (0 of 10)

## 7. Maintenance & community, 38 out of 100, up to 5.4 more on the total

Why it scored 38: The newest dated entry is the dots release note for 5 to 9 October 2026, three days before the check (30). Dots first appears in the DevDay note of 29 September 2026 and the weekly digest for 28 September to 2 October, which describe the same launch, so we count two dated entries, not three (0). A weekly public changelog and a community forum linked from the docs. The help centre refused our reader (8 of 15). No SDK, package or registry entry for dots (0). No package to assess (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 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.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 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.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## What we couldn't check

What 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.

- Kind is `platform`. Dots is a hosted assistant used inside ChatGPT, Slack and Teams with no API of its own, which the briefs define as a desktop, web or phone app. It is graded as a hosted service.
- The lead's address was right and its description matched the docs. Dots is documented inside the Codex docs set and delegates to Codex and ChatGPT Work. We list it apart from `openai-codex` because it has its own docs section, admin guide, access rules and release notes. The owner may prefer to treat it as part of ChatGPT.
- Capability `agent.mcp-client` is not claimed. Dots uses ChatGPT plugins, and we did not read the plugin docs to establish whether a user can add an MCP server.
- unchecked: the terms and privacy policy that govern Dots. No page read names them, https://openai.com/policies answered 403, and `provenance.terms` and `provenance.privacy` are left out. OpenAI's other listings link https://openai.com/policies/services-agreement/ and https://openai.com/policies/privacy-policy/, which were not read. Consumer Terms of Use may govern Pro accounts.
- unchecked: a status page and its incident record. None is linked from the pages read, so Reliability scores both lines as absent. A linked status page with a clean record would add up to 50.
- unchecked: published security advisories and incidents, the disclosure policy, bounty and certifications. openai.com was closed to our reader, so the deduction of 0 means none was found on pages we could read, not that none exists.
- unchecked: chatgpt.com, including chatgpt.com/dots, chatgpt.com/pricing and the plan pages. Its robots.txt disallows anthropic-ai and Claude-Web from every path. The Business Premium price and the Pro plan comparison are unread.
- unchecked: four help centre articles linked from the docs (getting started with your dot, data controls, Pro plans, app permissions). help.openai.com answered each with a JavaScript check. We requested all four in one command before reading the first answer, which the brief forbids after a 403.
- unchecked: where app tokens, saved website passwords and the dot's notes are stored, how long notes are kept, and what the deletion confirmation says happens to conversations, memories and scheduled tasks.
- unchecked: the legal entity. No page read names it.
- The cloud security page and whitepaper describe ChatGPT Work Cloud. They name dots only for the shared cloud browser, network and local-access controls, so retention figures taken from them may not apply to dots.
- Maintenance counts the DevDay note and the weekly digest of the same week as one entry. Counting them separately would add 20.
- We sent 16 requests to learn.chatgpt.com, one over the brief's guide of about fifteen.

## Weaknesses

- Beta. The admin setting is labelled Use dots (Beta), access is rolling out gradually, and Teams access is an invite-only alpha
- No way to view or delete individual saved notes is documented. The admin guide's link to reset options lands on the section for deleting the whole dot
- Custom rules are described as instructions the dot tries to follow and can get wrong, and an authorised write can run without a prompt where policy allows
- Pro consumer accounts in the European Economic Area, the United Kingdom and Switzerland are excluded, and the plan's allowance for deeper work has no published figure
- No status page, service terms or privacy policy is linked from the Dots docs, and openai.com and help.openai.com refused our reader
- No API, MCP endpoint or other documented way for an outside program to message a dot was found

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Say what may be sent. Asking a dot to draft replies does not permit sending them, and an action outside a stated scope needs another decision
- Stop three things separately. Pause stops the main task only, delegated tasks stop in Activity and recurring tasks in Scheduled, and completed actions are not undone
- Connect each thing separately. A messaging channel grants no access to an inbox, an app or a computer
- Sign out of websites and disconnect apps yourself when removing access. Revoking dots access or disconnecting an app does not delete information already obtained
- Keep the connected computer online with the ChatGPT app open for local work. Only one personal computer can be connected at a time

## When it's done

Send 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.