DataHub by Acryl Data, Inc. (DataHub)

# Fix list: DataHub

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/datahub, the October 2026 research run, assessed 3 October 2026. Grade C, 59.5 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 DataHub: 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. Payments & pricing, 10 out of 100, up to 11.3 more on the total

Why it scored 10: Scored on the paid option, as the self-hosted rule asks for open-source software sold beside a hosted version (DataHub Cloud). No x402, MPP or L402 (0). Cloud pricing is contact sales and datahub.com/pricing/ returns 404 (0). DataHub Core is free to self-host with no card, but Cloud has no free tier or trial we could find (10 of 20). A person deploys DataHub and creates a token in the UI, or signs a Cloud contract (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.

## 2. Reliability, 68 out of 100, up to 6.4 more on the total

Why it scored 68: Read with the local-software lines, since what an agent runs is mcp-server-datahub against a DataHub instance its owner hosts (the same server works against DataHub Cloud). PyPI and Docker packages, Python 3.11+ stated, but no minimum DataHub version is published. The server hides tools a too-old GMS can't run, and issue #64 asking for the minimum has been open since December 2025 (15 of 20). ci.yml runs lint, unit tests, Docker checks and integration tests against DataHub quickstarts for Cloud, v1.3.0 and head, and the last three runs on main passed (25). 12 open issues, four of them bug reports (#110, #118, #131, #139) open since April to July, and we couldn't see replies. DataHub itself has 521 open issues and 781 open pull requests (15 of 25). The changelog follows Keep a Changelog and marks breaking changes, but stops at 0.5.3 while tags reach 0.7.1, a breaking HTTP change sits under Unreleased, and 0.5.3 replaced the `filters` object with a `filter` string while calling it not breaking. DataHub's Updating DataHub page has a Breaking Changes section for every release (8 of 15). DataHub is at 1.7 with an LTS line, but the MCP server agents load is 0.7.1 (5 of 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.

## 3. Security & auth, 65 out of 100, up to 6.1 more on the total

Why it scored 65: Personal access tokens per user, gated by a privilege, expiring after 1 hour to 365 days with never-expiring tokens off by default, and revocable. HTTP mode takes each user's own bearer token in the header and rejects `?access_token=`, and Cloud's managed endpoint uses OAuth with dynamic client registration. Tokens have no scopes of their own and carry the user's full privileges (26 of 30). Write tools are off by default, save_document may only update documents the agent created unless the operator changes that, DataHub policies and views narrow what a user sees, but there's no confirmation step on writes (14 of 20). Tools return descriptions, documents and SQL that people and source systems wrote. The server strips HTML and base64 embeds but publishes no injection guidance (4 of 15). DataHub records the acting user on metadata changes and the docs say per-user tokens keep audit attribution per user, and we found no per-call MCP log for the operator (8 of 15). SECURITY.md with security@datahub.com and a PGP key, 10 GitHub advisories with four in the last twelve months, SOC 2 claimed for Cloud and third-party penetration tests on the security page. No bug bounty and no security.txt (13 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.

## 4. Agent ergonomics, 78 out of 100, up to 3.6 more on the total

Why it scored 78: Eight tools by default, within the ten the rubric names, but they carry about 26,000 characters of descriptions (roughly 6,500 tokens), because search and get_lineage each embed the same 3,063-character filter grammar. Write, user and data-quality tools sit behind environment flags, which work as toolsets (18 of 25). `offset` and `num_results` (max 50), facet-only search, keyword filters on schema fields, lineage hops and an 80,000-token response budget that truncates long descriptions (20). Errors come back as tool errors naming the bad input and the next step, without codes (15 of 20). The 10 read tools carry readOnlyHint, the 12 write tools carry no annotations at all, and there are no idempotency keys (10 of 20). Every read tool runs on defaults (`query="*"`), and DataHub ships Python and Java SDKs (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.

## 5. Schema & documentation, 81 out of 100, up to 3.1 more on the total

Why it scored 81: Every tool has a JSON Schema generated from Python type hints, and each instance serves an OpenAPI spec and GraphQL (25). llms.txt at docs.datahub.com (10). Descriptions say what each tool is for and when to reach for another (list_schema_fields "when search results truncate fields"), with a full filter grammar. The docs page lists tools the open-source server doesn't register and annotations it doesn't set (15 of 20). Literal enums for direction, sort order, query source and description operation, but filters are one free string with a grammar in prose, the 50-result cap lives only in prose, and properties carry no descriptions in the schema (9 of 15). Worked examples in every description and errors that say what to do next ("use the search tool ... or create the tags first"), but no error catalogue (12 of 15). Tags and PyPI versions, and DataHub's release notes, but the MCP changelog is five releases behind (10 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.

## 6. Transparency & trust, 65 out of 100, up to 3.1 more on the total

Made of editorial 55, provenance 75.

Why it scored 65: DataHub Core and the MCP server are Apache-2.0. DataHub Cloud and its extra MCP tools are closed (27 of 30). The privacy policy of 11 August 2026 excludes paying customers' use of the service, the terms of 28 May 2020 cover the website, and we found no DPA, subprocessor list or Cloud retention statement. The security page names AWS and KMS encryption (6 of 30). Updating DataHub lists Breaking Changes, Deprecations and Potential Downtime for every release, and v1.6.0 has an LTS line, but there's no stated notice period (12 of 20). Telemetry is on by default and documented with an opt-out for the server and the CLI, but the MCP server's per-call events and error text aren't on the telemetry page, the README or the MCP docs (10 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: datahub.com, no registry record we could read (0 of 15)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 77 out of 100, up to 2 more on the total

Why it scored 77: DataHub v1.6.0.3 tagged on 25 September 2026 and mcp-server-datahub 0.7.1 on PyPI on 16 September (30). DataHub v1.7.0, v1.6.0.1, v1.6.0.2, v1.7.0.1 and v1.6.0.3 since 4 August, plus MCP server 0.7.0 and 0.7.1 (20). The MCP server has 12 open issues, the oldest from December 2025, and DataHub has 521 open issues and 781 open pull requests with a stale-issue workflow. We couldn't see reply times (12 of 25). Not in the official MCP registry, where a search returns third-party DataHub servers. Official Python and Java SDKs are current (7 of 15). CI on every push. Dependabot arrived in the MCP repository on 14 September 2026 in the same change that patched 40 open dependency advisories (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.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-03-11. Since the modular rewrite synced on 11 March 2026, every mcp-server-datahub tool call sends a Mixpanel event, on by default, with the tool name, the MCP client's name and version (added 16 March), result length, duration and the first 500 characters of any error message. The telemetry page lists CLI invocations and error types, and neither it, the README nor the MCP docs mention MCP tool calls. Error text can carry URNs from the catalogue, so more than counts, -3 (https://github.com/acryldata/mcp-server-datahub/blob/main/src/mcp_server_datahub/_telemetry.py; https://github.com/datahub-project/datahub/blob/master/docs/deploy/telemetry.md)
- 2026-02-04 to 2026-05-19. Four advisories in twelve months, CVE-2026-25644 (7.5, the LDAP ingestion source turned off TLS certificate checks, fixed in 1.3.1.8), CVE-2026-44501 (4.3, cookie deserialisation in the OIDC callback, fixed in 1.5.0.3) and two open redirects. All fixed and published, so decayed, -2 (https://github.com/datahub-project/datahub/security/advisories)

## 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.

- unchecked: DataHub Cloud prices, contract terms, DPA and subprocessors, which sit behind sales and a Master Services Agreement.
- unchecked: who answers issues on either repository and how fast. GitHub's issue pages showed no comment counts to our reader.
- unchecked: whether the official MCP registry lists the acryldata server under another name. A search for datahub returned only third-party servers, and the repository has no server.json.
- unchecked: GitHub release dates for the DataHub tags. We used the commit dates of v1.6.0.1 to v1.6.0.3 and v1.7.0.1, while v1.7.0 has a release page dated 4 August.
- The SOC 2 claim is on the Cloud vs Core page only. The security page doesn't mention it and we found no trust centre.

## Weaknesses

- The eight default tools carry about 26,000 characters of descriptions, and search and get_lineage each repeat the same 3,063-character filter grammar
- Every tool call sends a Mixpanel event by default with the tool name, the client and up to 500 characters of any error message, and no docs page mentions it
- The MCP server is pre-1.0 (0.7.1), and CHANGELOG.md stops at 0.5.3 with a breaking HTTP change still under Unreleased
- The docs page lists Cloud-only tools such as find_sql_context and says every tool carries destructiveHint and idempotentHint, which the open-source server doesn't set
- No published DataHub Cloud prices or trial, no DPA or subprocessor list, and the privacy policy excludes paying customers' use of the service

## 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.

- Start keyword queries with `/q` and pass filters as one string, such as `entity_type = dataset AND platform = snowflake`
- Call `search` with `num_results=0` first to get the tags, glossary terms, platforms and domains in use
- Page with `offset`. `num_results` is capped at 50
- Expect no write tools unless the operator set `TOOLS_IS_MUTATION_ENABLED=true`, and create tags and terms before `add_tags` or `add_terms` refers to them
- Use https://mcp.datahub.com/mcp with OAuth only on DataHub Cloud v1.0.2 or later. DataHub Core needs the local server and a personal access token

## What the review panel asked for

- mark Cloud-only tools
- publish minimum version
- document MCP telemetry
- per-token scopes

## 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.