OpenSearch

# Fix list: OpenSearch

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/opensearch, the October 2026 research run, assessed 9 October 2026. Grade B, 68 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 OpenSearch: 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, 70 out of 100, up to 5.3 more on the total

Why it scored 70: API keys since 3.7 carry cluster and index permissions, expire within 90 days at most and are revoked on all nodes at once. They are off until enabled, and only a security administrator can create them. Basic auth, client certificates, JWT, OpenID Connect and SAML also work (27 of 30). Roles with document-level and field-level security allow read-only access to named indexes. The MCP server allows writes by default and has no confirmation step, though `OPENSEARCH_SETTINGS_ALLOW_WRITE=false` blocks write methods (14 of 20). Search returns stored documents as written, and no prompt-injection guidance was found. The MCP server has an opt-in guard that restricts caller-named URLs to public addresses (3 of 15). Audit logs with configurable categories, an `API_TOKEN_WRITE` category for keys, and structured JSON logs of every MCP tool call (14 of 15). `SECURITY.md` with a reporting address, published advisories, CodeQL in CI and a stated 60-day target for fixing known vulnerabilities. No security.txt (HTTP 404), and no bug bounty or certification found (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. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Read with the self-hosted rule. Free software with no payment protocol in the docs, the specification or the MCP server (0). Nothing to buy from the foundation, so no price to publish (20). No card or trial needed (20). An agent can pull the Docker image and call the API with no signup (20).

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.

## 3. Reliability, 77 out of 100, up to 4.6 more on the total

Why it scored 77: Read with the local-software lines, because the foundation runs no hosted service and the owner runs the cluster. No status page or SLA exists for the software. Docker image `opensearchproject/opensearch`, tarball, RPM and Debian packages with a bundled JDK (20). The core repository holds 24 public workflows, among them `gradle-check.yml`, `codeql-analysis.yml` and `detect-breaking-change.yml`. Whether the default branch passes was not checked (15 of 25). GitHub shows about 2,500 open issues, and 21 of the 25 newest carried the `untriaged` label on 9 October 2026, most of them one to three days old (12 of 25). The releases page states semantic versioning with breaking changes only in major versions, and a breaking-changes page lists them per version (15). Version 3.9.0, with 1.0 in July 2021 (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.

## 4. Transparency & trust, 57 out of 100, up to 3.8 more on the total

Made of editorial 69, provenance 45.

Why it scored 57: Apache-2.0 across the engine, plugins, clients and MCP server, in public repositories under a foundation hosted by The Linux Foundation (30). The software runs on the owner's machines and the data stays there. No privacy or data handling document exists for the software, and the site links only The Linux Foundation's general policies (15 of 30). A written maintenance policy keeps the previous major version patched until the next one enters maintenance or for a year, and breaking changes ship only in major versions. No notice period for removing a single API was found (16 of 20). No statement on telemetry was found in the install or about pages read. The MCP server adds a `User-Agent` naming itself to requests it sends to the cluster (8 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):

- Endpoint on the vendor's domain:  is not on opensearch.org (0 of 15)
- Terms of service: not found (0 of 10)
- Privacy policy: not found (0 of 10)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 5. Schema & documentation, 77 out of 100, up to 3.7 more on the total

Why it scored 77: OpenAPI 3.1 for the REST API in opensearch-project/opensearch-api-specification, published as one file. We counted about 717 operations in 36 namespace files in the source (25). No `llms.txt` on docs.opensearch.org (HTTP 404). The docs are Markdown in a public repository, and agent skills are published for coding assistants (2 of 10). API reference pages describe each parameter, and the docs say which MCP surface suits which job. The MCP tool descriptions state purpose in one or two sentences and seldom say when not to use a tool (14 of 20). Parameters are typed in the specification, with enums such as `op_type`. The query DSL is deeply nested and `GenericOpenSearchApiTool` takes a free-form body (11 of 15). Request and response examples on most reference pages. No catalogue of error types was found in the pages read (10 of 15). A version history with dated release notes, a breaking-changes page and a changelog for the specification (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, 80 out of 100, up to 3.3 more on the total

Why it scored 80: Responses are sized with `size`, `_source` and `filter_path`. The MCP server starts with nine core tools and keeps four more categories off until enabled, and hides connection fields from tool schemas once a cluster is configured (22 of 25). `from` and `size`, `search_after`, point in time, and filters inside or outside the k-NN clause (20). Errors come back as JSON with a type and reason, 429 is documented for concurrency limits, and the MCP server sets `isError` since 0.11.0. No error catalogue (13 of 20). `op_type=create`, `if_seq_no` and `if_primary_term` make writes safe to repeat. No MCP tool sets `readOnlyHint` or `destructiveHint` in the source at 0.12.0 (12 of 20). Official clients in eight languages. A vector index needs a mapping before any search (13 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.

## 7. Maintenance & community, 84 out of 100, up to 1.4 more on the total

Why it scored 84: OpenSearch 3.9.0 was tagged on 22 September 2026, 17 days before the check (30). Core releases 3.8.0 on 31 July and 3.9.0 on 22 September, and MCP server releases 0.11.0 on 28 July and 0.12.0 on 17 September, fall inside 90 days. The core alone has two (20). 200 commits reached the core's main branch between 18 August and 9 October 2026, and new issues are labelled within days, but about 2,500 stay open. Reply times were not sampled (15 of 25). Official clients in eight languages are listed in the docs. Their current versions and the MCP registry were not checked (10 of 15). Dependabot, CodeQL and a breaking-change detector run in the core repository, and the MCP server pins patched `aiohttp` and `click` versions (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.

- Until 0.12.0 (fix merged 3 August 2026, released 17 September 2026) the OpenSearch MCP Server attached its own configured credentials to any cluster URL a caller named in dynamic connection mode, which had shipped in 0.10.0 on 2 June 2026. The fix is described in the changelog and commit, and the repository lists no published advisory. Three points (https://github.com/opensearch-project/opensearch-mcp-server-py/blob/main/CHANGELOG.md).
- The Security plugin published four advisories on 30 April 2026, one moderate (GHSA-x83w-23jp-g6pw, document-level security not applied to documents linked by `has_child` or `has_parent`) and three low. Fixed and disclosed, so two points (https://github.com/opensearch-project/security/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: whether the core's `gradle-check` and other workflows pass on the default branch
- unchecked: PyPI and npm download counts, and the current versions of the official clients
- unchecked: whether the MCP server is in the official MCP registry. No `server.json` is in its repository
- unchecked: issue reply times in the core repository. Only the newest page of open issues was read
- unchecked: the published OpenAPI file at api-spec.opensearch.org. The count of about 717 operations comes from the YAML source and a pattern match, so it is approximate
- unchecked: where https://opensearch.org/llms.txt redirects. It answered 301 and was not followed
- Whether OpenSearch or OpenSearch Dashboards sends any usage telemetry. No statement was found in the pages read
- The date 3.9.0 was announced to the public. `lastRelease` is the tag date, 22 September 2026, and the schedule's window closed on 29 September
- The maintenance table on the releases page still names 3.7.0 as the latest minor version of the 3 line
- The docs page for the MCP server asks for Python 3.11 or later, while the package's `pyproject.toml` says 3.10 or later
- The lead held. The docs site has no `llms.txt`, and terms and privacy are left unset because the foundation publishes none for the software

## Weaknesses

- No hosted service from the foundation. The owner installs, secures and runs the cluster, and no status page or SLA exists for the software
- No `llms.txt` on docs.opensearch.org (HTTP 404), and doc pages carry about 160 KB of navigation text each
- The MCP server allows writes by default (`OPENSEARCH_SETTINGS_ALLOW_WRITE` is `true`) and no tool sets `readOnlyHint` or `destructiveHint`
- Before version 0.12.0 the MCP server sent its own configured credentials to any cluster URL a caller named. No advisory was published
- The core repository shows about 2,500 open issues, and 21 of the 25 newest carried the `untriaged` label on 9 October 2026

## 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.

- Create the index with `index.knn` set to true and a `knn_vector` field of the right dimension before indexing vectors
- Put the `filter` clause inside the `knn` query for filtering during the search. A filter outside it runs afterwards and can return fewer than `k` hits
- Ask a security administrator for an API key and send it as `Authorization: ApiKey <token>`. The token is shown once and lasts 90 days at most
- Use `size`, `_source` and `filter_path` to keep responses small, and `search_after` with a point in time for deep paging
- Set `OPENSEARCH_SETTINGS_ALLOW_WRITE=false` on the MCP server unless writes are needed. `GenericOpenSearchApiTool` can call any endpoint

## 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.