Be the first to get access to new connectors. Get early access →
ConnectorBay Early access
Discover
OpenAI evaluations and runs (BYOK) icon

OpenAI evaluations and runs (BYOK)

Read your OpenAI evaluations, their runs, pass and fail counts and per-item results metadata with your own API key

Unify Dynamics LLC

Tools5read-only
Sign-inBYOKbring your own key
Key keptNeverused for one call only
PriceFreeoptional Pro limits
ProviderOpenAI Evalsunofficial, not affiliated
Works withMCP · RESTany agent

Try asking

Real questions, and the tools your agent calls to answer them.

Which evaluations have I defined in my OpenAI project?
Tools called
list_evals
Show the details of OpenAI evaluation eval_abc123
Tools called
get_eval
How did the runs of OpenAI evaluation eval_abc123 turn out?
Tools called
list_eval_runs
What were the pass and fail counts of OpenAI eval run evalrun_abc123?
Tools called
get_eval_run
Which items failed in OpenAI eval run evalrun_abc123?
Tools called
list_eval_run_output_items

About

OpenAI evaluations and runs (BYOK) lets an AI assistant read your own OpenAI Evals data through the OpenAI Evals API with a key you supply (OpenAI API key). Read-only: 5 tools (list_evals, get_eval, list_eval_runs, get_eval_run, list_eval_run_output_items). The BYOK key travels with each request, is used for that one call only and is never stored, logged or cached; it is only ever sent to api.openai.com. Unofficial connector by Unify Dynamics LLC: not affiliated with, endorsed by or sponsored by OpenAI Evals. OpenAI Evals and its logos are trademarks of their owners and are used only to say which service the connector works with.

Sources: Live from the OpenAI Evals API (api.openai.com) with a BYOK key; documentation: https://platform.openai.com/docs/api-reference/evals. Endpoints checked against that documentation on 2026-10-16.

Privacy

The publisher, Unify Dynamics LLC, states that tool inputs are processed in memory and not retained.

Data not collectedNo account, no profile, no stored prompts or results.

What it can do

5 tools, grouped by job. Every one is read-only or compute-only.

OpenAPI
Read your data5 tools
  • List evalsList the evaluations of the key's project with name and creation time, ordered by creation or update time.
  • Get evalGet one evaluation: name, creation time, metadata, and the data source configuration type.
  • List eval runsList the runs of one evaluation with status, model, report link and pass, fail and error counts.
  • Get eval runGet one evaluation run: status, model, creation time, report link, pass, fail and error counts, and any error.
  • List eval run output itemsList the output items of an evaluation run with pass or fail status and ids (the item contents are not picked).

Information

PublisherUnify Dynamics LLC
CategoryDeveloper tools
LanguagesEnglish
Built forWorldwide
CompatibilityAny MCP client · any agent · REST
Version0.1.0
MCP endpointhttps://byok-openai-evals.connectorbay.com/mcp
REST APIopenapi.json
SupportGet help · Status
CountryWorldwide
FunctionOther
LayerDeep connectors

Connect it to your agent

Three ways in, the same 5 tools.

MCP (Claude, Cursor, Windsurf, VS Code, any MCP client)https://byok-openai-evals.connectorbay.com/mcp
RESTPOST https://byok-openai-evals.connectorbay.com/v1/tools/<tool>
OpenAI function callinghttps://connectorbay.com/openai/byok-openai-evals.json
OpenAI-compatible chathttps://connectorbay.com/v1/chat/completions · how it works
  • Add the MCP URL in your client's connector, MCP or integration settings (streamable HTTP, no OAuth).
  • Key: send your OpenAI API key in the X-Upstream-Key header (get one at platform.openai.com).
  • Ask a question; the agent picks the right tool and answers from the live data.

OpenAI evaluations and runs connector: questions and answers

What is the OpenAI evaluations and runs connector?

OpenAI evaluations and runs (BYOK) lets an AI assistant read your own OpenAI Evals data through the OpenAI Evals API with a key you supply (OpenAI… It is a hosted MCP server with 5 tools such as List evals, Get eval, List eval runs, Get eval run, List eval run output items, published by Unify Dynamics LLC. It is independent and not affiliated with OpenAI Evals.

How do I connect OpenAI evaluations and runs to Claude, ChatGPT or Cursor?

Add the MCP endpoint https://byok-openai-evals.connectorbay.com/mcp to your client's MCP, connector or integration settings (streamable HTTP, no OAuth). For REST, POST JSON to https://byok-openai-evals.connectorbay.com/v1/tools/<tool>; the OpenAPI description is at /openapi.json. For OpenAI function calling, load https://connectorbay.com/openai/byok-openai-evals.json.

Do I need a OpenAI Evals account or API key?

Yes: this is a bring-your-own-key connector. Create a OpenAI API key at https://platform.openai.com/api-keys and send it in the X-Upstream-Key header. ConnectorBay uses it for that one call only and never stores, logs or caches it.

Is OpenAI evaluations and runs read-only?

Yes. Every tool reads or computes. Nothing is created, changed, sent, paid or deleted in your account.

Is the OpenAI evaluations and runs connector free?

Yes. It is free to use within the published rate limits; an optional Pro key raises them. There is no account to create.

More like this

More in Other

Mobilize (BYOK)

Read volunteer events, time slots and organizations from Mobilize with your own organization API key

Live · MCP + REST7 tools

Mode (BYOK)

Read your Mode workspace collections, reports, report runs and queries with your own Mode API token and secret (paid Mode plans)

Live · MCP + REST8 tools

Modrinth (BYOK)

Search Modrinth mods, plugins and other projects, read versions and authors, and your own account data with your own personal access token

Live · MCP + REST10 tools

Modrinth Reference (BYOK)

Read Modrinth category, loader and game version lists, licence texts, random projects, dependencies and look up projects, versions and files in bulk with your own personal access token

Live · MCP + REST11 tools

Moesif (BYOK)

Read your Moesif API analytics applications, governance rules, billing meters, plans, prices and usage reports with your own management API token

Live · MCP + REST7 tools

Monte Carlo (BYOK)

Read your Monte Carlo data observability tables, alerts, monitors, domains, data products and roles with your own API key ID and token (GraphQL queries only)

Live · MCP + REST8 tools

Documentation

Back to top

OpenAI evaluations and runs (BYOK) — connector documentation

  • MCP endpoint: https://byok-openai-evals.connectorbay.com/mcp
  • REST API: POST https://byok-openai-evals.connectorbay.com/v1/tools/{tool_name}, described by the OpenAPI 3.1 spec at https://byok-openai-evals.connectorbay.com/openapi.json
  • Transport: Streamable HTTP, stateless, MCP specification 2025-06-18 or later
  • Authentication: your own OpenAI Evals OpenAI API key, sent in the X-Upstream-Key header of every request (never stored, logged or cached)
  • Health: https://byok-openai-evals.connectorbay.com/health · Tool schemas: https://byok-openai-evals.connectorbay.com/tools.json
  • Publisher: Unify Dynamics LLC · connectorbay.com/byok-openai-evals · agent@connectorbay.com

What OpenAI evaluations and runs (BYOK) is for

OpenAI evaluations and runs (BYOK) lets an AI assistant read your own OpenAI Evals data through the OpenAI Evals API with a key you supply (OpenAI API key). Read-only: 5 tools (list_evals, get_eval, list_eval_runs, get_eval_run, list_eval_run_output_items). The BYOK key travels with each request, is used for that one call only and is never stored, logged or cached; it is only ever sent to api.openai.com. Unofficial connector by Unify Dynamics LLC: not affiliated with, endorsed by or sponsored by OpenAI Evals. OpenAI Evals and its logos are trademarks of their owners and are used only to say which service the connector works with.

Ask your assistant things like:

  • "Which evaluations have I defined in my OpenAI project?"
  • "Show the details of OpenAI evaluation eval_abc123"
  • "How did the runs of OpenAI evaluation eval_abc123 turn out?"
  • "What were the pass and fail counts of OpenAI eval run evalrun_abc123?"
  • "Which items failed in OpenAI eval run evalrun_abc123?"

What it does

  • list_evals: List the evaluations of the key's project with name and creation time, ordered by creation or update time. Reads evaluation metadata only: it starts no run and sends no prompt.
  • get_eval: Get one evaluation: name, creation time, metadata, and the data source configuration type. Reads evaluation metadata only: it starts no run and sends no prompt.
  • list_eval_runs: List the runs of one evaluation with status, model, report link and pass, fail and error counts. Reads evaluation metadata only: it starts no run and sends no prompt.
  • get_eval_run: Get one evaluation run: status, model, creation time, report link, pass, fail and error counts, and any error. Reads evaluation metadata only: it starts no run and sends no prompt.
  • list_eval_run_output_items: List the output items of an evaluation run with pass or fail status and ids (the item contents are not picked). Reads evaluation metadata only: it starts no run and sends no prompt.

BYOK key

You bring your own OpenAI Evals OpenAI API key (https://platform.openai.com/api-keys). Your assistant sends it in the X-Upstream-Key header with each request. ConnectorBay uses it for that one call only: it is never stored, logged or cached, never shown in an answer, and it is only sent to api.openai.com over https, with no redirects followed. Create a key with the fewest permissions the tools need (read-only if the provider offers it), and revoke it at the provider whenever you like.

Data sources

Live from the OpenAI Evals API (api.openai.com) with a BYOK key; documentation: https://platform.openai.com/docs/api-reference/evals. Endpoints checked against that documentation on 2026-10-16.

What it is not

Unofficial connector by Unify Dynamics LLC: not affiliated with, endorsed by or sponsored by OpenAI Evals. OpenAI Evals and its logos are trademarks of their owners and are used only to say which service the connector works with. It is read-only: it never creates, changes, sends or deletes anything in your OpenAI Evals account.

Built for: Worldwide. Languages: English.

Privacy policy

Back to top

OpenAI evaluations and runs (BYOK) — Privacy Policy

Effective 24 October 2026. Unify Dynamics LLC ("we", "us"), Menlo Park, California.

What OpenAI evaluations and runs (BYOK) is. OpenAI evaluations and runs (BYOK) is a connector (an MCP server and REST API) that an AI assistant calls on your behalf to read your own OpenAI Evals data: Read your OpenAI evaluations, their runs, pass and fail counts and per-item results metadata with your own API key. You bring your own OpenAI Evals OpenAI API key; the assistant sends it to us with each request.

BYOK key. The key you supply is used to make the one OpenAI Evals call you asked for. Our application does not store it, write it to a log, put it in a cache, or share it. It is sent only to api.openai.com, over https, and is removed from any text we return to you or your assistant. We do not follow redirects with your key. If OpenAI Evals rejects it, we say so without repeating it.

What we receive. The arguments of each tool call, the BYOK key (for that call only), the technical headers needed to serve the request (an IP address or a client identifier supplied by the assistant) and standard server logs. What OpenAI Evals receives. The request we make on your behalf, including the key and the arguments, under OpenAI Evals's own terms and privacy policy. OpenAI Evals acts at your direction and is the only recipient of the data of your request besides our hosting provider.

What we return. The answer OpenAI Evals gives for your request, trimmed to the fields the tool documents. We do not keep it: it is returned to your assistant and not stored, cached or used to train anything.

What we do not do. We do not build a profile, sell or share personal information, or keep your OpenAI Evals data. Rate-limit counters (a client identifier and timestamps) are held in memory for at most 24 hours. Hosting access logs (IP addresses, request paths and status codes) are kept for up to 30 days for security and abuse prevention; our application adds no request headers or bodies to them.

Sub-processors. OpenAI Evals (the API you call with a BYOK key, at your direction). Hosting: Vercel (United States regions).

Your rights. Because we store no personal data beyond short-lived logs, requests to access, correct or delete data are usually satisfied automatically; you can still contact agent@connectorbay.com and we will respond within 30 days. Revoke your key at OpenAI Evals at any time to end all access.

Children. OpenAI evaluations and runs (BYOK) is a general-audience service operated by the adult using the assistant, under OpenAI Evals's own age rules.

Changes. We will post changes here with a new effective date.

Terms of service

Back to top

OpenAI evaluations and runs (BYOK) — Terms of Service

Effective 24 October 2026. These terms govern use of the OpenAI evaluations and runs (BYOK) connector ("the Service") operated by Unify Dynamics LLC ("we", "us").

  1. The Service. OpenAI evaluations and runs (BYOK) provides: Read your OpenAI evaluations, their runs, pass and fail counts and per-item results metadata with your own API key. It is accessed through an AI assistant or directly over MCP or HTTP, and is read-only.
  2. Your own key, your own account. You must use your own OpenAI Evals credentials and only access data you are allowed to access. OpenAI Evals's terms apply to your use of OpenAI Evals; you are responsible for complying with them, including any rate limits and permitted uses. Do not share a key you do not own.
  3. Independent. Unofficial connector by Unify Dynamics LLC: not affiliated with, endorsed by or sponsored by OpenAI Evals. OpenAI Evals and its logos are trademarks of their owners and are used only to say which service the connector works with.
  4. No advice, no guarantees. Output is the data OpenAI Evals returns, lightly trimmed; it may be incomplete, delayed or changed by OpenAI Evals. Check important facts at the source.
  5. Read-only. The Service does not create, change, send or delete anything in your OpenAI Evals account.
  6. Acceptable use. Automated use within the published rate limits is welcome. Do not use the Service to access data you are not entitled to, to circumvent OpenAI Evals's controls, or to attack any system.
  7. Availability. The Service is provided "as is" without warranty. We may change, limit or discontinue it at any time; OpenAI Evals may change or end its API at any time.
  8. Data. Our Privacy Policy describes what we process. In short: the BYOK key is used for the one call and not stored or logged.
  9. Pro keys. Optional ConnectorBay Pro API keys are issued at our discretion and are separate from your OpenAI Evals key.
  10. Liability. To the fullest extent permitted by law, Unify Dynamics LLC is not liable for any loss arising from use of or reliance on the Service, including any action OpenAI Evals takes on your account. Our total liability is limited to US $100.
  11. Governing law. California law; venue in San Mateo County, California.
  12. Contact. Unify Dynamics LLC · agent@connectorbay.com · https://connectorbay.com/byok-openai-evals

OpenAI evaluations and runs (BYOK) support

Email agent@connectorbay.com. We answer within two business days.

When you report a problem, include the question you asked your agent, roughly when, and which agent you used (for example Claude, ChatGPT or Gemini). Never send us passwords, account numbers or ID numbers. OpenAI evaluations and runs (BYOK) never asks for them.

Get OpenAI evaluations and runs (BYOK)

Free · BYOK · Read-only

BYOK: bring your own keyThis connector uses your own OpenAI Evals OpenAI API key. Create one at https://platform.openai.com/api-keys, with the fewest permissions it needs (read-only if the provider offers it). Your assistant sends it in the X-Upstream-Key header with every request. It is used for that one call only: never stored, logged or cached, and only ever sent to OpenAI Evals. Revoke it at the provider whenever you like. Unofficial connector: not affiliated with OpenAI Evals.