# Fix list: Presidio From Anchor Terminal's listing at https://www.anchorterminal.com/tools/microsoft-presidio, the October 2026 research run, assessed 8 October 2026. Grade B, 66 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 Presidio: 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, 53 out of 100, up to 8.2 more on the total Why it scored 53: Local-software reading. The REST services have no authentication, and the FAQ says this is by design and that a gateway, proxy or service mesh should sit in front (10 of 30). The analyser only reads text and returns spans. The anonymiser returns a transformed copy and deletes nothing. There are no roles or approval steps (12 of 20). It returns positions, types and scores for the caller's own text, not content from third parties (10). Requests accept a correlation_id that is written to the service log, and return_decision_process explains each match. No audit log beyond that (9 of 15). SECURITY.md routes reports to GitHub private vulnerability reporting with acknowledgement in 48 hours, CodeQL runs in CI and workflow actions are pinned to commit hashes. No published advisories, no bug bounty and no security.txt (12 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, 69 out of 100, up to 5 more on the total Why it scored 69: API reading for the REST services. An OpenAPI 3.0 document covers eight paths of the analyser and anonymiser. The image redactor's /redact is not in it (22 of 25). presidio.dataprivacystack.org/llms.txt returns 404 (0). Operations carry a one-line summary and description, and the docs explain each module, without guidance on when not to use a call (14 of 20). Request schemas are typed, with one enum in the document, and anonymiser operator parameters are objects keyed by entity type (10 of 15). Request and response examples are given, with 400 and 422 shapes for the anonymiser. The analyser's 500 on missing text is not in the document (11 of 15). Tagged releases and a public changelog, though 2.2.364 has no changelog section and the document's own version reads 2.0 (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, 69 out of 100, up to 5 more on the total Why it scored 69: `/analyze` takes an entity list, a score threshold and an allow list, and leaves the decision trace off unless return_decision_process is set, so responses can be kept small (20 of 25). Text can be a string or an array for batches. There is no pagination, and none is needed for span lists (14 of 20). Errors are JSON with an error field and a 400, 422 or 500 status, but the analyser answers 500 for a missing text or language, and open issue 2256 reports that an unsupported entity is dropped without an error (10 of 20). Calls are stateless and safe to retry. Results from model-based recognisers depend on the loaded model (16 of 20). Two required fields and working defaults, with a Python library only (9 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. Payments & pricing, 60 out of 100, up to 5 more on the total Why it scored 60: MIT-licensed package the owner runs, with no hosted or paid option from the project, read with the self-hosted rule. 20 for pricing, 20 for a free start and 20 for use without an account or key. No x402, MPP or L402 in the docs or source (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. ## 5. Reliability, 78 out of 100, up to 4.4 more on the total Why it scored 78: Local-software reading. Installs from PyPI as `presidio-analyzer` and `presidio-anonymizer`, with Python 3.10 to 3.14 stated in the install docs and package metadata (20). Public CI on GitHub Actions tests each package on five Python versions. Six of the last eight runs on main passed, one was cancelled and the run of 8 October 2026 failed (20 of 25). 68 open issues, two of them labelled bug, and 98 open pull requests. Most issues opened since 20 September have no reply (15 of 25). Versions count up as 2.2.x for every release, feature releases included, and CHANGELOG.md has no section for 2.2.364, whose changes sit under unreleased (8 of 15). Version 2.2.364, past 1.0, though the image redactor package is 0.0.60 (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. ## 6. Maintenance & community, 63 out of 100, up to 3.2 more on the total Why it scored 63: 2.2.364 was released on 22 July 2026, 78 days before the check (20). It is the only release in the last 90 days, with 2.2.363 on 28 June just outside (5 of 20, a quarter of the line where earlier dossiers gave half for two). 30 issues were opened since 1 August and 11 are closed, main had 14 commits in the 30 days to 8 October, and 98 pull requests are open (15 of 25). Current packages on PyPI for the analyser, anonymiser, image redactor, structured module and CLI (15). Dependabot, committed uv lockfiles and CI on five Python versions, with the latest run on main failing (8 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. ## 7. Transparency & trust, 65 out of 100, up to 3.1 more on the total Made of editorial 76, provenance 53. Why it scored 65: MIT licence, with the copyright line changed from Microsoft Corporation to Presidio Contributors in July 2026 (30). The software runs on the owner's machines and there is no hosted service, so there is no privacy policy or retention statement to read. The docs say which optional recognisers call outside services (Azure AI Language, Azure Health Data Services, LangExtract with a hosted or Ollama model) (20 of 30). Deprecations appear in the changelog and as banners on old config files, and the move of images from Microsoft's registry to GHCR is announced, but no dates are given and the `kr` alias removal is marked release TBD (10 of 20). We found no telemetry code in the repository and no statement on telemetry in the docs (16 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: dataprivacystack.org, registered 2026-04-13 (under a year) (0 of 15) - Status page: not found (0 of 10) - security.txt: not found (0 of 10) ## 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. - The lead named Microsoft as vendor. Presidio left Microsoft for the community organisation Data Privacy Stack (announced 29 June 2026). github.com/microsoft/presidio and microsoft.github.io/presidio redirect to the new addresses. The slug keeps the old name. - No legal entity is named for Data Privacy Stack on its site, the docs or the licence. - No privacy policy is published, so provenance.privacy is left out. The MIT licence stands in for terms. - Whether the project runs no telemetry is inferred from a search of the source. The docs make no statement. - unchecked: the governance documents in the data-privacy-stack/MVG repository were not read. - unchecked: issue reply times were read from comment counts in the GitHub search API, not from each thread. ## Weaknesses - The REST containers have no authentication by design. The FAQ says to put a gateway or proxy in front - SUPPORT.md states no SLA and no official support. The project is run by volunteers since leaving Microsoft - One release in the 90 days to 8 October 2026 (2.2.364 on 22 July), and CHANGELOG.md has no section for it - Covers personal data only. No prompt injection, jailbreak or content moderation checks - The README warns that detection is automated and may miss personal data, so other protections are still needed - 98 open pull requests, and most issues opened since 20 September 2026 had no reply on 8 October ## 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. - Install from PyPI or pull images from ghcr.io/data-privacy-stack. The mcr.microsoft.com/presidio-* images are no longer updated - Download a spaCy model (python -m spacy download en_core_web_lg) before the first `AnalyzerEngine()` call, or use the Docker image - Send both text and language to `/analyze`. A request missing either returns HTTP 500 with a JSON error field - Pass entities and score_threshold to limit results. Many country-specific recognisers are disabled by default and need enabling in the registry YAML - Keep the containers on a private network or behind your own authenticating proxy. They accept any caller ## 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.