# Fix list: GPT4All From Anchor Terminal's listing at https://www.anchorterminal.com/tools/gpt4all, the October 2026 research run, assessed 3 October 2026. Grade F, 36.3 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 GPT4All: 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. Security & auth, 28 out of 100, up to 12.6 more on the total Why it scored 28: Read with the tool checklist. The local API server has no authentication, listens on 127.0.0.1:4891 and is off by default. Every response carries `Access-Control-Allow-Origin: *` (server.cpp line 614), so any web page open in the user's browser can call it while it's on (issue #3681), and the host-header check against DNS rebinding sits unmerged on a branch from 27 May 2025. Keys for remote providers and the Nomic Embed API are kept in the app's own files, not a system keychain (5 of 30). The API has no write actions, the server is opt-in, and LocalDocs can only be switched on from the app, but there's no read-only key or per-caller control, because callers can't be told apart (10 of 20). LocalDocs puts the text of the owner's files into prompts and returns it through the API, and we found no prompt-injection guidance (5 of 15). API calls show in the app's server chat, with no log beyond it (5 of 15). No SECURITY.md, no security.txt at nomic.ai (404) and no advisories, and the two security reports of 26 June 2026 are public issues with no reply. The model catalogue, release notes and fallback downloads use plain HTTP, and nine request sites set `QSslSocket::VerifyNone`, among them model downloads and the analytics and Datalake uploads. The macOS build is signed and CI checks the signature (3 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. ## 2. Schema & documentation, 40 out of 100, up to 9.8 more on the total Why it scored 40: Read with the API lines for the local server, the one surface an agent can call. No OpenAPI or other contract. The server mirrors part of OpenAI's API, and the Python SDK has typed signatures and a reference page (5 of 25). No llms.txt (docs.gpt4all.io/llms.txt returns the home page) (0). The API server page lists four endpoints in one table and says the server is HTTP on 127.0.0.1 only, but doesn't say that `stream`, `tools`, `response_format` and `functions` are refused (8 of 20). The server checks the type of every OpenAI parameter and names the ones it won't take, and the SDK takes typed arguments with defaults (8 of 15). curl, PowerShell and Python examples, with no documented errors (7 of 15). A dated changelog for the app and the SDK, though the SDK's Unreleased section has waited since 2024 (12 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. ## 3. Agent ergonomics, 41 out of 100, up to 9.6 more on the total Why it scored 41: Read with the tool lines against the local API server, which a person has to switch on in Settings. `max_tokens` and `n` size a response, there's no streaming, and LocalDocs references come back attached to each answer with no control over how many (12 of 25). Four endpoints, with no paging or filters (6 of 20). Errors use OpenAI's envelope with a message naming the refused parameter, such as "'tools' is not supported", but none are documented (12 of 20). No idempotency keys or retry guidance. Completions are stateless, and API calls run through a single server chat in the app (6 of 20). No tool calling, structured output or streaming, so an agent can't use it for tool use, and the only clients are the Python SDK (August 2024) and a TypeScript binding (5 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. ## 4. Reliability, 56 out of 100, up to 8.8 more on the total Why it scored 56: Read with the local-software lines, since GPT4All is a desktop app the owner runs, with no hosted service behind its API. Installers for Windows x64 and ARM64, macOS 12.6 or later and x86-64 Linux, with minimum and recommended hardware published, a Flathub build the README calls community-maintained, and the Python SDK on PyPI (2.8.2, Python 3.8 or later) (18 of 20). CircleCI builds, signs and test-installs the app and runs the binding tests, but the last commit to main is from 27 May 2025 and the pipeline pages need JavaScript our reader lacks, so whether it passes today is unknown (8 of 25). 729 open issues and 42 open pull requests. The 2026 reports we read, among them download stalls (#3692), a Qwen3.5 report labelled bug-unconfirmed (#3657) and two security reports, had no maintainer reply we could see (3 of 25). A Keep a Changelog file with dated Added, Changed and Fixed sections per version, and semver tags (12 of 15). 3.10.0 (15). 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. ## 5. Maintenance & community, 6 out of 100, up to 8.2 more on the total Why it scored 6: v3.10.0, dated 24 February 2025 in the changelog, 586 days before this check (0). No release in the last 90 days (0). No commit on main since 27 May 2025, 729 open issues and 42 open pull requests, and no maintainer reply we could see on the 2026 issues we opened, including one of 9 July asking whether development has stopped (#3690) and the two security reports (2 of 25). The Python SDK's last release is 2.8.2 of 14 August 2024, with changes waiting in an Unreleased section, and there's no MCP server or registry entry (3 of 15). The llama.cpp fork and CI haven't moved since May 2025 (1 of 10). 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. ## 6. Payments & pricing, 60 out of 100, up to 5 more on the total Why it scored 60: Read with the self-hosted rule. No x402, MPP or L402, and Nomic sells nothing for GPT4All (0). MIT with no account and no card, so 20, 20 and 20 on the last three lines. Remote models bill the user's own OpenAI, Groq or Mistral key. 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. ## 7. Transparency & trust, 57 out of 100, up to 3.8 more on the total Made of editorial 54, provenance 59. Why it scored 57: MIT for the app, the backend and the bindings, copyright Nomic, Inc. (30). The docs call GPT4All completely private, and the product page says no data leaves your machine. The app fetches its model catalogue, release notes and news over plain HTTP at start, and Nomic's privacy policy (15 January 2026) covers its website and platform, names Mixpanel and US servers, gives no retention periods and doesn't mention GPT4All or the Datalake. The opt-in text says shared chats become public and attribution is kept (10 of 30). No deprecation policy, and nothing from Nomic says whether GPT4All is still maintained (2 of 20). Analytics and the Datalake are opt-in at first start, off by default, with the terms shown. A user who declines still sends one opt_out event with a random device ID on first start and after each upgrade, the analytics described as anonymous add the user's IP from api.ipify.org, and the settings docs list only the Datalake toggle (12 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): - Domain age: nomic.ai, no registry record we could read (0 of 15) - Status page: not found (0 of 10) - security.txt: not found (0 of 10) ## Deductions Each comes off the total. A fixed and documented problem counts for less at the next check. - 2026-06-26. Two security reports filed as public issues, unanswered, with no release since. #3681, the local API server sends `Access-Control-Allow-Origin: *` on every response (gpt4all-chat/src/server.cpp line 614) and has no authentication, so a web page open in the user's browser can call /v1/chat/completions while the server is on and read the answers, including LocalDocs snippets from the owner's files. #3682, the model catalogue and the fallback model download use plain http://gpt4all.io, and the app turns TLS certificate checks off on nine request sites. A host-header fix has sat unmerged on the mitigate-dns-rebind branch since 27 May 2025. Unfixed, with the server off by default, -6. https://github.com/nomic-ai/gpt4all/issues/3681; https://github.com/nomic-ai/gpt4all/issues/3682 ## 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. - Whether Nomic still maintains GPT4All. Nothing on nomic.ai or GitHub says it has stopped, and issue #3690 asking the question had no Nomic reply we could see - unchecked: whether CircleCI passes on main today, since its pages need JavaScript - unchecked: the state and npm downloads of the TypeScript binding - unchecked: the registration dates of nomic.ai and gpt4all.io - unchecked: who replied to the 2026 issues and when, since comment threads didn't fully render for our reader ## Weaknesses - No release since 24 February 2025 and no commit to main since 27 May 2025 - The local server has no authentication and sends `Access-Control-Allow-Origin: *`, so a web page can call it while it's on - The model catalogue and fallback downloads use plain HTTP, and nine request sites turn TLS certificate checks off - No streaming, tool calling or structured output on the API - No SECURITY.md or security.txt, and the June 2026 security reports have no reply ## 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. - Ask the owner to tick Enable Local API Server in Settings. Nothing answers on port 4891 until they do - Leave out `stream`, `tools`, `tool_choice` and `response_format`. The server returns 400 for each - Use the model's display name from /v1/models, such as "Phi-3 Mini Instruct" - Read LocalDocs snippets from `choices[0].references`. Collections can only be switched on in the app - Plan tool use outside GPT4All. Its API can't call tools ## What the review panel asked for - a maintenance statement - a dated sunset notice - merge the DNS-rebinding fix - turn certificate checks back on ## 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.