1. You can now add streamable http servers via the CLI
2. As part of this, I'm also changing the existing bearer_token plain
text config field with ane env var
```
mcp add github --url https://api.githubcopilot.com/mcp/ --bearer-token-env-var=GITHUB_PAT
```
This lets users/companies explicitly choose whether to force/disallow
the keyring/fallback file storage for mcp credentials.
People who develop with Codex will want to use this until we sign
binaries or else each ad-hoc debug builds will require keychain access
on every build. I don't love this and am open to other ideas for how to
handle that.
```toml
mcp_oauth_credentials_store = "auto"
mcp_oauth_credentials_store = "file"
mcp_oauth_credentials_store = "keyrung"
```
Defaults to `auto`
# External (non-OpenAI) Pull Request Requirements
Before opening this Pull Request, please read the dedicated
"Contributing" markdown file or your PR may be closed:
https://github.com/openai/codex/blob/main/docs/contributing.md
If your PR conforms to our contribution guidelines, replace this text
with a detailed and high quality description of your changes.
## Summary
- ensure the TypeScript SDK sets CODEX_INTERNAL_ORIGINATOR_OVERRIDE to
codex_sdk_ts when spawning the Codex CLI
- extend the responses proxy test helper to capture request headers for
assertions
- add coverage that verifies Codex threads launched from the TypeScript
SDK send the codex_sdk_ts originator header
## Testing
- Not Run (not requested)
------
https://chatgpt.com/codex/tasks/task_i_68e561b125248320a487f129093d16e7
We use to put the review prompt in the first user message as well to
bypass statsig overrides, but now that's been resolved and instructions
are being respected, so we're duplicating the review instructions.
this fixes an issue where text lines with long words would sometimes
overflow.
- the default penalties for the OptimalFit algorithm allow overflowing
in some cases. this seems insane to me, and i considered just banning
the OptimalFit algorithm by disabling the 'smawk' feature on textwrap,
but decided to keep it and just bump the overflow penalty to ~infinity
since optimal fit does sometimes produce nicer wrappings. it's not clear
this is worth it, though, and maybe we should just dump the optimal fit
algorithm completely.
- user history messages weren't rendering with the same wrap algorithm
as used in the composer, which sometimes resulted in wrapping messages
differently in the history vs. in the composer.
Before this change:
```
tamird@L03G26TD12 codex-rs % codex
zsh: do you wish to see all 3864 possibilities (1285 lines)?
```
After this change:
```
tamird@L03G26TD12 codex-rs % codex
app-server -- [experimental] Run the app server
apply a -- Apply the latest diff produced by Codex agent as a `git apply` to your local working tree
cloud -- [EXPERIMENTAL] Browse tasks from Codex Cloud and apply changes locally
completion -- Generate shell completion scripts
debug -- Internal debugging commands
exec e -- Run Codex non-interactively
generate-ts -- Internal: generate TypeScript protocol bindings
help -- Print this message or the help of the given subcommand(s)
login -- Manage login
logout -- Remove stored authentication credentials
mcp -- [experimental] Run Codex as an MCP server and manage MCP servers
mcp-server -- [experimental] Run the Codex MCP server (stdio transport)
responses-api-proxy -- Internal: run the responses API proxy
resume -- Resume a previous interactive session (picker by default; use --last to continue the most recent)
```
when exiting a session that was started with `codex resume`, the note
about how to resume again wasn't being printed.
thanks @aibrahim-oai for pointing out this issue!
`http_config.auth_header` automatically added `Bearer `. By adding it
ourselves, we were sending `Bearer Bearer <token>`.
I confirmed that the GitHub MCP initialization 400s before and works
now.
I also optimized the oauth flow to not check the keyring if you
explicitly pass in a bearer token.
## Summary
- add a `codex sandbox` subcommand with macOS and Linux targets while
keeping the legacy `codex debug` aliases
- update documentation to highlight the new sandbox entrypoints and
point existing references to the new command
- clarify the core README about the linux sandbox helper alias
## Testing
- just fmt
- just fix -p codex-cli
- cargo test -p codex-cli
------
https://chatgpt.com/codex/tasks/task_i_68e2e00ca1e8832d8bff53aa0b50b49e
## Summary
- replace manual wiremock SSE mounts in the compact suite with the
shared response helpers
- simplify the exec auth_env integration test by using the
mount_sse_once_match helper
- rely on mount_sse_sequence plus server request collection to replace
the bespoke SeqResponder utility in tests
## Testing
- just fmt
------
https://chatgpt.com/codex/tasks/task_i_68e2e238f2a88320a337f0b9e4098093
## Summary
- add a reusable `ev_response_created` helper that builds
`response.created` SSE events for integration tests
- update the exec and core integration suites to use the new helper
instead of repeating manual JSON literals
- keep the streaming fixtures consistent by relying on the shared helper
in every touched test
## Testing
- `just fmt`
------
https://chatgpt.com/codex/tasks/task_i_68e1fe885bb883208aafffb94218da61
`codex-responses-api-proxy` is designed so that there should be exactly
one copy of the API key in memory (that is `mlock`'d on UNIX), but in
practice, I was seeing two when I dumped the process data from
`/proc/$PID/mem`.
It appears that `std::io::stdin()` maintains an internal `BufReader`
that we cannot zero out, so this PR changes the implementation on UNIX
so that we use a low-level `read(2)` instead.
Even though it seems like it would be incredibly unlikely, we also make
this logic tolerant of short reads. Either `\n` or `EOF` must be sent to
signal the end of the key written to stdin.
## Summary
- replace manual event polling loops in several core test suites with
the shared wait_for_event helpers
- keep prior assertions intact by using closure captures for stateful
expectations, including plan updates, patch lifecycles, and review flow
checks
- rely on wait_for_event_with_timeout where longer waits are required,
simplifying timeout handling
## Testing
- just fmt
------
https://chatgpt.com/codex/tasks/task_i_68e1d58582d483208febadc5f90dd95e
## Summary
This PR is an alternative approach to #4711, but instead of changing our
storage, parses out shell calls in the client and reserializes them on
the fly before we send them out as part of the request.
What this changes:
1. Adds additional serialization logic when the
ApplyPatchToolType::Freeform is in use.
2. Adds a --custom-apply-patch flag to enable this setting on a
session-by-session basis.
This change is delicate, but is not meant to be permanent. It is meant
to be the first step in a migration:
1. (This PR) Add in-flight serialization with config
2. Update model_family default
3. Update serialization logic to store turn outputs in a structured
format, with logic to serialize based on model_family setting.
4. Remove this rewrite in-flight logic.
## Test Plan
- [x] Additional unit tests added
- [x] Integration tests added
- [x] Tested locally
In the past, we were treating `input exceeded context window` as a
streaming error and retrying on it. Retrying on it has no point because
it won't change the behavior. In this PR, we surface the error to the
client without retry and also send a token count event to indicate that
the context window is full.
<img width="650" height="125" alt="image"
src="https://github.com/user-attachments/assets/c26b1213-4c27-4bfc-90f4-51a270a3efd5"
/>
We truncate the output of exec commands to not blow the context window.
However, some cases we weren't doing that. This caused reports of people
with 76% context window left facing `input exceeded context window`
which is weird.
The `experimental_use_rmcp_client` flag is still useful to:
1. Toggle between stdio clients
2. Enable oauth beacuse we want to land
https://github.com/modelcontextprotocol/rust-sdk/pull/469,
https://github.com/openai/codex/pull/4677, and binary signing before we
enable it by default
However, for no-auth http servers, there is only one option so we don't
need the flag and it seems to be working pretty well.
Codex isn’t great yet on Windows outside of WSL, and while we’ve merged
https://github.com/openai/codex/pull/4269 to reduce the repetitive
manual approvals on readonly commands, we’ve noticed that users seem to
have more issues with GPT-5-Codex than with GPT-5 on Windows.
This change makes GPT-5 the default for Windows users while we continue
to improve the CLI harness and model for GPT-5-Codex on Windows.
## Summary
- Factor `load_config_as_toml` into `core::config_loader` so config
loading is reusable across callers.
- Layer `~/.codex/config.toml`, optional `~/.codex/managed_config.toml`,
and macOS managed preferences (base64) with recursive table merging and
scoped threads per source.
## Config Flow
```
Managed prefs (macOS profile: com.openai.codex/config_toml_base64)
▲
│
~/.codex/managed_config.toml │ (optional file-based override)
▲
│
~/.codex/config.toml (user-defined settings)
```
- The loader searches under the resolved `CODEX_HOME` directory
(defaults to `~/.codex`).
- Managed configs let administrators ship fleet-wide overrides via
device profiles which is useful for enforcing certain settings like
sandbox or approval defaults.
- For nested hash tables: overlays merge recursively. Child tables are
merged key-by-key, while scalar or array values replace the prior layer
entirely. This lets admins add or tweak individual fields without
clobbering unrelated user settings.
This PR adds oauth login support to streamable http servers when
`experimental_use_rmcp_client` is enabled.
This PR is large but represents the minimal amount of work required for
this to work. To keep this PR smaller, login can only be done with
`codex mcp login` and `codex mcp logout` but it doesn't appear in `/mcp`
or `codex mcp list` yet. Fingers crossed that this is the last large MCP
PR and that subsequent PRs can be smaller.
Under the hood, credentials are stored using platform credential
managers using the [keyring crate](https://crates.io/crates/keyring).
When the keyring isn't available, it falls back to storing credentials
in `CODEX_HOME/.credentials.json` which is consistent with how other
coding agents handle authentication.
I tested this on macOS, Windows, WSL (ubuntu), and Linux. I wasn't able
to test the dbus store on linux but did verify that the fallback works.
One quirk is that if you have credentials, during development, every
build will have its own ad-hoc binary so the keyring won't recognize the
reader as being the same as the write so it may ask for the user's
password. I may add an override to disable this or allow
users/enterprises to opt-out of the keyring storage if it causes issues.
<img width="5064" height="686" alt="CleanShot 2025-09-30 at 19 31 40"
src="https://github.com/user-attachments/assets/9573f9b4-07f1-4160-83b8-2920db287e2d"
/>
<img width="745" height="486" alt="image"
src="https://github.com/user-attachments/assets/9562649b-ea5f-4f22-ace2-d0cb438b143e"
/>
