Open data for Codex
Codex data.
Ready to build with.
Not affiliated with OpenAI. An independent, open mirror and dataset for the OpenAI Codex client.
The official model catalog, verified feature flags, the ModelInfo schema and hook product data for developers building on Codex. Open JSON endpoints, immutable snapshots, and a traceable source for every dataset.
Data at a glance
- Mirror health
- See health endpoint
- Last successful check
- See metadata
- Content fetched
- See metadata
- Source plan
- See metadata
- Schema tag
- Verified tags
- Feature registry tag
- Verified tags
Catalog checks and content updates are separate. Schema and feature versions follow verified Codex tags. Health comes from /healthz; missing metadata is not a healthy status.
START BUILDING
Quick start
Fetch the model catalog as JSON. Use its ETag with
If-None-Match for subsequent requests.
curl -fsS https://data.cp.dev/v1/codex/models.json
API reference / v1
The catalog is mirrored live from chatgpt.com/backend-api/codex/models and
served only at data.cp.dev. Every other dataset is a static file generated from
the sources in the repository, so it can be mirrored anywhere.
| Endpoint | Description |
|---|---|
/v1/index.json
|
Discovery: endpoints, last sync summary, licenses. |
/v1/codex/models.json
|
Latest official Codex catalog envelope {"models":[…]}, unmodified.
ETag + If-None-Match supported.
|
/v1/codex/meta.json
|
When it was fetched/checked, client version used, content hash, source plan label. |
/v1/codex/snapshots/index.json
|
Recent immutable snapshots; each at
/v1/codex/snapshots/<sha256>.json.
|
/v1/codex/changes.json
|
Models added / removed / changed, newest first. |
/v1/schema/codex-model-info/latest.json
|
JSON Schema of the Codex client's ModelInfo rejection rules; per verified
tag at /<tag>.json, tag list at /index.json. Static
file, updated per verified tag.
|
/v1/features/codex/latest.json
|
Feature-flag registry per verified Codex tag: key, lifecycle stage, default, official
rustdoc + /experimental menu copy, legacy aliases (e.g. telepathy →
chronicle), cross-tag history, and community-curated annotations (zh
today; translations welcome). Per tag at /<tag>.json, tag list at
/index.json. Static file, updated per verified tag; field reference in
docs/DATASETS.md.
|
/v1/hooks/codex/latest.json
|
Hook products, path matching rules, purpose and removal notes. Contribute a product. |
/v1/compat/codex/latest.json
|
Compatibility intelligence for the Codex Pass client: which managed
config.toml keys each Codex release still accepts, plus upgrade
advisories. Probed automatically on every release; maintained by the Codex Pass team
and not open for contribution.
|
/healthz
|
200 after a successful catalog check within 24 hours and with no permanent sync failure; 503 otherwise. |
Source & freshness
-
The Codex catalog reflects one dedicated ChatGPT account's view (plan label in
meta.json); other plans may see a different list. -
It is fetched with the latest published
@openai/codexclient version. Older clients ignore unknown fields. - If the upstream is unreachable, the last successfully fetched catalog keeps being served.
Licenses
Code: MIT. Datasets (feature-flag registry + annotations, ModelInfo schema, compat, hook
registry): CC-BY-4.0, derived from openai/codex sources (Apache-2.0, reproduced with LICENSE
+ NOTICE in the repository). Product marks in the hook registry retain their owners' rights.
/v1/codex/* bodies are served as-is from OpenAI; CodexData grants no license
over them.