OpenAI Agents SDK by OpenAI

# Fix list: OpenAI Agents SDK

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/openai-agents-sdk, the October 2026 research run, assessed 1 October 2026. Grade AA, 86.5 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 OpenAI Agents SDK: 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, 60 out of 100, up to 5 more on the total

Why it scored 60: No payment protocol (0). Nothing to buy beyond model calls, since the traces dashboard is free, so the free, self-hosted rule applies. The MIT package is public and free (20), needs no card (20) and no account, and runs non-OpenAI and local models (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.

## 2. Security & auth, 80 out of 100, up to 3.5 more on the total

Why it scored 80: Tracing is on by default and trace_include_sensitive_data defaults to true, so model and function-call inputs and outputs go to OpenAI's Traces dashboard. OPENAI_AGENTS_DISABLE_TRACING=1, set_tracing_disabled or RunConfig turn it off (10). Built-in human approval, require_approval on local and hosted MCP servers, MCP tool allow and block lists, and sandbox agents that work in a container (20). Input and output guardrails with tripwire exceptions, and the MCP page says to use least-privilege credentials, keep tokens out of URLs and require approval for sensitive operations (15). Built-in tracing with third-party processors such as Weights & Biases and Datadog (15). SECURITY.md routes reports through OpenAI's coordinated disclosure policy, which governs bug bounty eligibility, and we found no advisories or CVEs against the SDK (20). Framework reading, so SOC 2 isn't scored.

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. Reliability, 85 out of 100, up to 3 more on the total

Why it scored 85: Official packages on PyPI (Python 3.10 or newer) and npm (20). The Tests workflow passes on main (25). 8 open issues and 3 open pull requests (25). A written policy for 0.Y.Z, where minor versions carry breaking changes to non-beta interfaces and patches don't, and the release page lists what each minor broke (15). 0.22.3, still pre-1.0 (0).

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. Schema & documentation, 95 out of 100, up to 0.8 more on the total

Why it scored 95: Typed Python API with a reference section in the docs (25). llms.txt, per the listing's earlier check (10). Guides cover hand-offs, agents as tools and code-driven orchestration, though we didn't re-check the when-not-to-use wording this run (15). Function tools get their schemas from typed Python signatures (15). Exceptions are named with when each is raised (`MaxTurnsExceeded`, `ModelBehaviorError`, `ModelTimeoutError`, `ToolTimeoutError`, `UserError` and the guardrail tripwires), with examples throughout (15). Versioning policy and release notes per minor (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.

## 5. Transparency & trust, 92 out of 100, up to 0.7 more on the total

Made of editorial 84, provenance 100.

Why it scored 92: MIT (30). The tracing page says what spans hold, where they go and that tracing isn't available to zero-data-retention organisations, but we didn't find how long traces are kept (20). Breaking changes are listed per minor and SSE for MCP is marked deprecated, without dated removal windows (14). Tracing and its content capture are disclosed with three ways to turn them off (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.

## 6. Agent ergonomics, 97 out of 100, up to 0.5 more on the total

Why it scored 97: An agent with one MCP server is about 11 lines, with static and dynamic tool filters and tool-list caching (25). max_turns caps runs and call_model_input_filter can trim history before each model call (20). Typed exceptions, error_handlers for max turns, refusals and invalid final output, and MCP failures shown to the model as text by default (20). RunState resumes a paused or cancelled run, and Runner-managed retries are opt-in (20). An agent needs a name and instructions, but 0.20.0 changed the default model (7). Python and JavaScript (5).

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.

## 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.

- We didn't find how long OpenAI keeps traces sent by the SDK
- We didn't check whether the JavaScript package states supported Node versions; its npm metadata has no engines field
- llms.txt and the API reference rest on the listing's check of 2026-09-26

## Weaknesses

- Tracing on by default, with model and tool content, sent to OpenAI
- Tracing isn't available to zero-data-retention organisations
- Breaking changes in each minor release while pre-1.0
- 0.20.0 changed the default model

## 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 OPENAI_AGENTS_DISABLE_TRACING=1, or OPENAI_AGENTS_TRACE_INCLUDE_SENSITIVE_DATA=0 to keep content out of traces
- Pin to a minor version. Each 0.Y can break
- Set require_approval on MCP servers that write
- Name the model explicitly. The default changed in 0.20.0
- Use an error handler for max_turns instead of catching MaxTurnsExceeded

## What the review panel asked for

- Make tracing opt-in
- State trace retention
- Tracing off by default
- Trace retention stated
- Token budget per run
- publish trace retention
- State the timeout defaults
- Say what a run does on a 429 without retries
- sensitive traces off
- published trace retention
- a dated removal for SSE
- Name a default model in the docs examples

## 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.