OpenMetadata by Collate, Inc.

# Fix list: OpenMetadata

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/openmetadata, the October 2026 research run, assessed 3 October 2026. Grade B, 66.9 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 OpenMetadata: 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, 20 out of 100, up to 10 more on the total

Why it scored 20: Scored on the paid option, as the self-hosted rule asks for open-source software sold beside a hosted version (Collate). No x402, MPP or L402 (0). The Free plan is published, but Premium, Enterprise and the AI add-on are contact sales (0). Collate Free needs no card and self-hosting is free, though Collate Free leaves MCP out (20). A person signs up in a browser or deploys an instance and creates a token (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. Security & auth, 65 out of 100, up to 6.1 more on the total

Why it scored 65: OAuth 2.0 with PKCE and dynamic client registration through the instance's SSO, 1-hour access tokens, rotating 7-day refresh tokens and a revocation endpoint, plus personal access tokens with configurable expiry and bot JWTs. Tokens carry the user's or bot's full roles with no scopes of their own (27 of 30). Calls run with the caller's roles and policies, three of the four write tools are marked destructive, but they're always listed, there's no read-only switch and no confirmation step (10 of 20). Tools return descriptions, knowledge pages, SQL that has run and context memories written by people and source systems, and we found no injection guidance (3 of 15). Every MCP call is recorded in the instance database with tool, user, outcome, latency and client, and OAuth operations are logged (12 of 15). SECURITY.md routes reports through GitHub advisories, 8 advisories are published with 3 in 2026, THREAT_MODEL.md and INCIDENT_RESPONSE.md sit in the repository, and Collate claims SOC 2 Type II. No security.txt or bug bounty, the supported-versions table is stale, and the open report on unmasked connections in the MCP entity tool (#34566) has no reply yet. We found no masking step in the MCP read path (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.

## 3. Transparency & trust, 55 out of 100, up to 3.9 more on the total

Made of editorial 51, provenance 59.

Why it scored 55: The platform is Apache-2.0, but the module an agent talks to has a `LICENSE` file with the `Collate Community License` 1.0, a source-available licence barring competing hosted services, while its Java headers and the README say Apache-2.0 (15 of 30). Collate's privacy policy (24 August 2022) covers the service, names Stripe and gives no retention period, and the terms (29 August 2022) host data in the United States, take a perpetual licence to contributions and set no deletion timeline. We found no DPA or subprocessor list, and the trust centre refused our reader (8 of 30). Breaking changes in release notes and versioned docs, but no deprecation notice policy and a stale supported-versions table (10 of 20). We found no telemetry to the vendor in the UI, ingestion or service dependencies, and UI web analytics post to the instance's own API, but no page says so either way (18 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: open-metadata.org, no registry record we could read (0 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 4. Reliability, 84 out of 100, up to 3.2 more on the total

Why it scored 84: Read with the local-software lines, since the MCP server agents use is part of the OpenMetadata server its owner runs, and Collate's hosted plans include MCP only on Enterprise. Docker images and Compose files per release, a Python ingestion package, and requirements stated (Docker 20.10+, Compose v2.1.1+, 6 GiB, 4 vCPUs) (20). 65 workflows, among them unit tests and integration tests on Postgres with OpenSearch and on MySQL with Elasticsearch, and the latest Postgres and OpenSearch integration runs on main passed (25). 688 open issues, dozens filed in the last two days, among them a semantic search bug in search_metadata (#34564) and an unmasked-connection report against the MCP entity tool (#34566), and we couldn't see replies (12 of 25). Release notes carry a breaking-changes section (2.0.0 moved semantic search settings and said search "stops working silently" without them), and patch lines run for 1.13.x and 2.0.x, but SECURITY.md still names 1.13.x and 1.12.x as supported (12 of 15). Version 2.0.3 (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. Schema & documentation, 85 out of 100, up to 2.4 more on the total

Why it scored 85: Every MCP tool has a typed JSON Schema in tools.json. The REST API has no public spec, since the openapi.json linked from llms.txt is a template titled OpenAPI Plant Store (22 of 25). llms.txt indexes about 3,000 pages across versions (10). Descriptions say when to use each tool, when to use another ("Choose this over semantic_search when the query targets known identifiers") and where answers go silently wrong (19 of 20). Required fields are marked, but few enums, `queryFilter` takes raw OpenSearch DSL in a string and patch_entity takes a JSON Patch blob (8 of 15). Examples inside the descriptions, and errors as JSON with a message and a status code, but no error catalogue in the docs (12 of 15). Versioned docs for 1.13.x and 2.0.x and release notes on every tag. The registry entry's 21 tools doesn't match the source (14 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. Agent ergonomics, 85 out of 100, up to 2.4 more on the total

Why it scored 85: 16 tools listed on a default deployment, with 13,653 characters of descriptions and 49,602 characters of definitions in all (about 12,400 tokens). The four RDF tools are withheld when RDF is off, which helps but isn't a subset the user picks (15 of 25). limit, offset and a `nextCursor` since 2.0, `fields` selection, column windows for wide tables, aggregations, persona search scopes and a response budget that flags `truncated` and `hasMore` (20). Errors come back with isError, a JSON message and a status code (400, 403 and so on), and soft errors inside a result are flagged too (18 of 20). All 20 definitions carry readOnlyHint and destructiveHint, with idempotentHint true on two, but there are no idempotency keys for the four write tools (17 of 20). Search and entity reads need few or no required fields, and the repository holds 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.

## 7. Maintenance & community, 89 out of 100, up to 1 more on the total

Why it scored 89: 2.0.3 tagged on 29 September 2026 (30). 1.12.14, 2.0.0-rc1, 1.13.4, 2.0.0, 2.0.1, 1.13.5, 1.13.6, 2.0.2 and 2.0.3 since 27 July (20). 688 open issues and 195 open pull requests, with labels on many of the newest reports, but we couldn't see reply times (15 of 25). Listed in the official MCP registry as io.github.open-metadata/openmetadata-mcp, a GitHub-verified namespace, since 10 September 2026, and Python and Java SDKs ship from the same repository (15). 65 CI workflows with integration tests passing on main (9 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.

- 2026-01-07 to 2026-05-14. Three advisories in 2026, all fixed and published. GHSA-5f29-2333-h9c7 (critical, 7 January, server-side template injection in FreeMarker email templates leading to code execution), CVE-2026-26010 (7.6, 11 February, bot JWTs exposed through /api/v1/ingestionPipelines to any read-only user, every version before 1.11.8) and CVE-2026-46481 (8.3, 14 May, TEST_CONNECTION returned the ingestion-bot JWT and a database password to non-admin users on 1.12.1, fixed in 1.12.4). Two exposed the bot tokens the docs suggest for unattended agents. Fixed, so decayed, -4 (https://github.com/open-metadata/OpenMetadata/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: Collate's trust centre (SOC 2 report, subprocessors, retention), which refused our reader.
- unchecked: whether the vendor confirms issue #34566 and how severe it rates it. The reporter says GHSA-c644-6j8x-5784 was filed first.
- unchecked: reply times on GitHub issues, which our reader showed without comment counts.
- unchecked: whether the OpenMetadata server sends any telemetry. We searched dependency manifests and the UI analytics path, not the whole service source.
- Which licence governs openmetadata-mcp. The module's `LICENSE` file is the `Collate Community License` 1.0, and its source headers and the repository README say Apache-2.0.
- Release dates are tag commit dates, except 2.0.0, whose release page we read.

## Weaknesses

- The 16 default tools take about 50,000 characters of definitions, 12,285 of them for search_metadata alone
- Three advisories in 2026, a critical template injection and two leaks of bot JWTs to low-privilege users
- An open report from 2 October 2026 says get_entity_details returns service connections (host, user, the stored password field) that the REST API masks, and we found no masking step in the MCP read path
- `openmetadata-mcp/LICENSE` is the `Collate Community License` 1.0, which bars competing hosted services, while the module's source headers and the README say Apache-2.0
- Write tools are always listed with no read-only switch, `queryFilter` takes raw OpenSearch DSL as a string, and the docs' openapi.json is a placeholder Plant Store spec

## 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.

- Set `entityType` for data-quality questions. testCase and testSuite are outside the default search scope
- Match a table's tests on `originEntityFQN`, not `entityFQN`. Column-level tests store the column's FQN there
- Read several known entities in one `search_metadata` call with a `terms` clause on `fullyQualifiedName` plus `fields`
- Use a bot JWT or a personal access token as `Authorization: Bearer` for unattended runs. OAuth needs a browser sign-in
- Call sparql_query, entity_neighborhood, find_by_tag or shacl_validate only if tools/list shows them. They appear only when the operator enabled RDF

## What the review panel asked for

- slimmer search_metadata
- real OpenAPI spec
- a read-only switch
- masking in MCP reads

## 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.