We had this annotation everywhere in app-server APIs which made it so
that fields get serialized as `field?: T`, meaning if the field as
`None` we would omit the field in the payload. Removing this annotation
changes it so that we return `field: T | null` instead, which makes
codex app-server's API more aligned with the convention of public OpenAI
APIs like Responses.
Separately, remove the `#[ts(optional_fields = nullable)]` annotations
that were recently added which made all the TS types become `field?: T |
null` which is not great since clients need to handle undefined and
null.
I think generally it'll be best to have optional types be either:
- `field: T | null` (preferred, aligned with public OpenAI APIs)
- `field?: T` where we have to, such as types generated from the MCP
schema:
https://github.com/modelcontextprotocol/modelcontextprotocol/blob/main/schema/2025-06-18/schema.ts
(see changes to `mcp-types/`)
I updated @etraut-openai's unit test to check that all generated TS
types are one or the other, not both (so will error if we have a type
that has `field?: T | null`). I don't think there's currently a good use
case for that - but we can always revisit.
Adds AgentMessageContentDelta, ReasoningContentDelta,
ReasoningRawContentDelta item streaming events while maintaining
compatibility for old events.
---------
Co-authored-by: Owen Lin <owen@openai.com>
In this PR, I am exploring migrating task kind to an invocation of
Codex. The main reason would be getting rid off multiple
`ConversationHistory` state and streamlining our context/history
management.
This approach depends on opening a channel between the sub-codex and
codex. This channel is responsible for forwarding `interactive`
(`approvals`) and `non-interactive` events. The `task` is responsible
for handling those events.
This opens the door for implementing `codex as a tool`, replacing
`compact` and `review`, and potentially subagents.
One consideration is this code is very similar to `app-server` specially
in the approval part. If in the future we wanted an interactive
`sub-codex` we should consider using `codex-mcp`
We currently have nested enums when sending raw response items in the
app-server protocol. This makes downstream schemas confusing because we
need to embed `type`-discriminated enums within each other.
This PR adds a small wrapper around the response item so we can keep the
schemas separate
This PR addresses a current hole in the TypeScript code generation for
the API server protocol. Fields that are marked as "Optional<>" in the
Rust code are serialized such that the value is omitted when it is
deserialized — appearing as `undefined`, but the TS type indicates
(incorrectly) that it is always defined but possibly `null`. This can
lead to subtle errors that the TypeScript compiler doesn't catch. The
fix is to include the `#[ts(optional_fields = nullable)]` macro for all
protocol structs that contain one or more `Optional<>` fields.
This PR also includes a new test that validates that all TS protocol
code containing "| null" in its type is marked optional ("?") to catch
cases where `#[ts(optional_fields = nullable)]` is omitted.
feature: Add "!cmd" user shell execution
This change lets users run local shell commands directly from the TUI by
prefixing their input with ! (e.g. !ls). Output is truncated to keep the
exec cell usable, and Ctrl-C cleanly
interrupts long-running commands (e.g. !sleep 10000).
**Summary of changes**
- Route Op::RunUserShellCommand through a dedicated UserShellCommandTask
(core/src/tasks/user_shell.rs), keeping the task logic out of codex.rs.
- Reuse the existing tool router: the task constructs a ToolCall for the
local_shell tool and relies on ShellHandler, so no manual MCP tool
lookup is required.
- Emit exec lifecycle events (ExecCommandBegin/ExecCommandEnd) so the
TUI can show command metadata, live output, and exit status.
**End-to-end flow**
**TUI handling**
1. ChatWidget::submit_user_message (TUI) intercepts messages starting
with !.
2. Non-empty commands dispatch Op::RunUserShellCommand { command };
empty commands surface a help hint.
3. No UserInput items are created, so nothing is enqueued for the model.
**Core submission loop**
4. The submission loop routes the op to handlers::run_user_shell_command
(core/src/codex.rs).
5. A fresh TurnContext is created and Session::spawn_user_shell_command
enqueues UserShellCommandTask.
**Task execution**
6. UserShellCommandTask::run emits TaskStartedEvent, formats the
command, and prepares a ToolCall targeting local_shell.
7. ToolCallRuntime::handle_tool_call dispatches to ShellHandler.
**Shell tool runtime**
8. ShellHandler::run_exec_like launches the process via the unified exec
runtime, honoring sandbox and shell policies, and emits
ExecCommandBegin/End.
9. Stdout/stderr are captured for the UI, but the task does not turn the
resulting ToolOutput into a model response.
**Completion**
10. After ExecCommandEnd, the task finishes without an assistant
message; the session marks it complete and the exec cell displays the
final output.
**Conversation context**
- The command and its output never enter the conversation history or the
model prompt; the flow is local-only.
- Only exec/task events are emitted for UI rendering.
**Demo video**
https://github.com/user-attachments/assets/fcd114b0-4304-4448-a367-a04c43e0b996
It's pretty amazing we have gotten here without the ability for the
model to see image content from MCP tool calls.
This PR builds off of 4391 and fixes#4819. I would like @KKcorps to get
adequete credit here but I also want to get this fix in ASAP so I gave
him a week to update it and haven't gotten a response so I'm going to
take it across the finish line.
This test highlights how absured the current situation is. I asked the
model to read this image using the Chrome MCP
<img width="2378" height="674" alt="image"
src="https://github.com/user-attachments/assets/9ef52608-72a2-4423-9f5e-7ae36b2b56e0"
/>
After this change, it correctly outputs:
> Captured the page: image dhows a dark terminal-style UI labeled
`OpenAI Codex (v0.0.0)` with prompt `model: gpt-5-codex medium` and
working directory `/codex/codex-rs`
(and more)
Before this change, it said:
> Took the full-page screenshot you asked for. It shows a long,
horizontally repeating pattern of stylized people in orange, light-blue,
and mustard clothing, holding hands in alternating poses against a white
background. No text or other graphics-just rows of flat illustration
stretching off to the right.
Without this change, the Figma, Playwright, Chrome, and other visual MCP
servers are pretty much entirely useless.
I tested this change with the openai respones api as well as a third
party completions api
Because conversations that use the Responses API can have encrypted
reasoning messages, trying to resume a conversation with a different
provider could lead to confusing "failed to decrypt" errors. (This is
reproducible by starting a conversation using ChatGPT login and resuming
it as a conversation that uses OpenAI models via Azure.)
This changes `ListConversationsParams` to take a `model_providers:
Option<Vec<String>>` and adds `model_provider` on each
`ConversationSummary` it returns so these cases can be disambiguated.
Note this ended up making changes to
`codex-rs/core/src/rollout/tests.rs` because it had a number of cases
where it expected `Some` for the value of `next_cursor`, but the list of
rollouts was complete, so according to this docstring:
bcd64c7e72/codex-rs/app-server-protocol/src/protocol.rs (L334-L337)
If there are no more items to return, then `next_cursor` should be
`None`. This PR updates that logic.
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/openai/codex/pull/5658).
* #5803
* #5793
* __->__ #5658
This PR adds support for a model-based summary and risk assessment for
commands that violate the sandbox policy and require user approval. This
aids the user in evaluating whether the command should be approved.
The feature works by taking a failed command and passing it back to the
model and asking it to summarize the command, give it a risk level (low,
medium, high) and a risk category (e.g. "data deletion" or "data
exfiltration"). It uses a new conversation thread so the context in the
existing thread doesn't influence the answer. If the call to the model
fails or takes longer than 5 seconds, it falls back to the current
behavior.
For now, this is an experimental feature and is gated by a config key
`experimental_sandbox_command_assessment`.
Here is a screen shot of the approval prompt showing the risk assessment
and summary.
<img width="723" height="282" alt="image"
src="https://github.com/user-attachments/assets/4597dd7c-d5a0-4e9f-9d13-414bd082fd6b"
/>
1. Adds AgentMessage, Reasoning, WebSearch items.
2. Switches the ResponseItem parsing to use new items and then also emit
3. Removes user-item kind and filters out "special" (environment) user
items when returning to clients.
Adds a `GET account/rateLimits/read` API to app-server. This calls the
codex backend to fetch the user's current rate limits.
This would be helpful in checking rate limits without having to send a
message.
For calling the codex backend usage API, I generated the types and
manually copied the relevant ones into `codex-backend-openapi-types`.
It'll be nice to extend our internal openapi generator to support Rust
so we don't have to run these manual steps.
# 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.
Adds a new ItemStarted event and delivers UserMessage as the first item
type (more to come).
Renames `InputItem` to `UserInput` considering we're using the `Item`
suffix for actual items.
The backend will be returning unix timestamps (seconds since epoch)
instead of RFC 3339 strings. This will make it more ergonomic for
developers to integrate against - no string parsing.
Add annotations and an export script that let us generate app-server
protocol types as typescript and JSONSchema.
The script itself is a bit hacky because we need to manually label some
of the types. Unfortunately it seems that enum variants don't get good
names by default and end up with something like `EventMsg1`,
`EventMsg2`, etc. I'm not an expert in this by any means, but since this
is only run manually and we already need to enumerate the types required
to describe the protocol, it didn't seem that much worse. An ideal
solution here would be to have some kind of root that we could generate
schemas for in one go, but I'm not sure if that's compatible with how we
generate the protocol today.
This change ensures that we store the absolute time instead of relative
offsets of when the primary and secondary rate limits will reset.
Previously these got recalculated relative to current time, which leads
to the displayed reset times to change over time, including after doing
a codex resume.
For previously changed sessions, this will cause the reset times to not
show due to this being a breaking change:
<img width="524" height="55" alt="Screenshot 2025-10-17 at 5 14 18 PM"
src="https://github.com/user-attachments/assets/53ebd43e-da25-4fef-9c47-94a529d40265"
/>
Fixes https://github.com/openai/codex/issues/4761
This adds `parsed_cmd: Vec<ParsedCommand>` to `ExecApprovalRequestEvent`
in the core protocol (`protocol/src/protocol.rs`), which is also what
this field is named on `ExecCommandBeginEvent`. Honestly, I don't love
the name (it sounds like a single command, but it is actually a list of
them), but I don't want to get distracted by a naming discussion right
now.
This also adds `parsed_cmd` to `ExecCommandApprovalParams` in
`codex-rs/app-server-protocol/src/protocol.rs`, so it will be available
via `codex app-server`, as well.
For consistency, I also updated `ExecApprovalElicitRequestParams` in
`codex-rs/mcp-server/src/exec_approval.rs` to include this field under
the name `codex_parsed_cmd`, as that struct already has a number of
special `codex_*` fields. Note this is the code for when Codex is used
as an MCP _server_ and therefore has to conform to the official spec for
an MCP elicitation type.
This adds a queryable auth status for MCP servers which is useful:
1. To determine whether a streamable HTTP server supports auth or not
based on whether or not it supports RFC 8414-3.2
2. Allow us to build a better user experience on top of MCP status
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"
/>
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"
/>
We continue the separation between `codex app-server` and `codex
mcp-server`.
In particular, we introduce a new crate, `codex-app-server-protocol`,
and migrate `codex-rs/protocol/src/mcp_protocol.rs` into it, renaming it
`codex-rs/app-server-protocol/src/protocol.rs`.
Because `ConversationId` was defined in `mcp_protocol.rs`, we move it
into its own file, `codex-rs/protocol/src/conversation_id.rs`, and
because it is referenced in a ton of places, we have to touch a lot of
files as part of this PR.
We also decide to get away from proper JSON-RPC 2.0 semantics, so we
also introduce `codex-rs/app-server-protocol/src/jsonrpc_lite.rs`, which
is basically the same `JSONRPCMessage` type defined in `mcp-types`
except with all of the `"jsonrpc": "2.0"` removed.
Getting rid of `"jsonrpc": "2.0"` makes our serialization logic
considerably simpler, as we can lean heavier on serde to serialize
directly into the wire format that we use now.
### Title
## otel
Codex can emit [OpenTelemetry](https://opentelemetry.io/) **log events**
that
describe each run: outbound API requests, streamed responses, user
input,
tool-approval decisions, and the result of every tool invocation. Export
is
**disabled by default** so local runs remain self-contained. Opt in by
adding an
`[otel]` table and choosing an exporter.
```toml
[otel]
environment = "staging" # defaults to "dev"
exporter = "none" # defaults to "none"; set to otlp-http or otlp-grpc to send events
log_user_prompt = false # defaults to false; redact prompt text unless explicitly enabled
```
Codex tags every exported event with `service.name = "codex-cli"`, the
CLI
version, and an `env` attribute so downstream collectors can distinguish
dev/staging/prod traffic. Only telemetry produced inside the
`codex_otel`
crate—the events listed below—is forwarded to the exporter.
### Event catalog
Every event shares a common set of metadata fields: `event.timestamp`,
`conversation.id`, `app.version`, `auth_mode` (when available),
`user.account_id` (when available), `terminal.type`, `model`, and
`slug`.
With OTEL enabled Codex emits the following event types (in addition to
the
metadata above):
- `codex.api_request`
- `cf_ray` (optional)
- `attempt`
- `duration_ms`
- `http.response.status_code` (optional)
- `error.message` (failures)
- `codex.sse_event`
- `event.kind`
- `duration_ms`
- `error.message` (failures)
- `input_token_count` (completion only)
- `output_token_count` (completion only)
- `cached_token_count` (completion only, optional)
- `reasoning_token_count` (completion only, optional)
- `tool_token_count` (completion only)
- `codex.user_prompt`
- `prompt_length`
- `prompt` (redacted unless `log_user_prompt = true`)
- `codex.tool_decision`
- `tool_name`
- `call_id`
- `decision` (`approved`, `approved_for_session`, `denied`, or `abort`)
- `source` (`config` or `user`)
- `codex.tool_result`
- `tool_name`
- `call_id`
- `arguments`
- `duration_ms` (execution time for the tool)
- `success` (`"true"` or `"false"`)
- `output`
### Choosing an exporter
Set `otel.exporter` to control where events go:
- `none` – leaves instrumentation active but skips exporting. This is
the
default.
- `otlp-http` – posts OTLP log records to an OTLP/HTTP collector.
Specify the
endpoint, protocol, and headers your collector expects:
```toml
[otel]
exporter = { otlp-http = {
endpoint = "https://otel.example.com/v1/logs",
protocol = "binary",
headers = { "x-otlp-api-key" = "${OTLP_TOKEN}" }
}}
```
- `otlp-grpc` – streams OTLP log records over gRPC. Provide the endpoint
and any
metadata headers:
```toml
[otel]
exporter = { otlp-grpc = {
endpoint = "https://otel.example.com:4317",
headers = { "x-otlp-meta" = "abc123" }
}}
```
If the exporter is `none` nothing is written anywhere; otherwise you
must run or point to your
own collector. All exporters run on a background batch worker that is
flushed on
shutdown.
If you build Codex from source the OTEL crate is still behind an `otel`
feature
flag; the official prebuilt binaries ship with the feature enabled. When
the
feature is disabled the telemetry hooks become no-ops so the CLI
continues to
function without the extra dependencies.
---------
Co-authored-by: Anton Panasenko <apanasenko@openai.com>
We currently get information about rate limits in the response headers.
We want to forward them to the clients to have better transparency.
UI/UX plans have been discussed and this information is needed.
We need to construct the history different when compact happens. For
this, we need to just consider the history after compact and convert
compact to a response item.
This needs to change and use `build_compact_history` when this #3446 is
merged.
## 📝 Review Mode -- Core
This PR introduces the Core implementation for Review mode:
- New op `Op::Review { prompt: String }:` spawns a child review task
with isolated context, a review‑specific system prompt, and a
`Config.review_model`.
- `EnteredReviewMode`: emitted when the child review session starts.
Every event from this point onwards reflects the review session.
- `ExitedReviewMode(Option<ReviewOutputEvent>)`: emitted when the review
finishes or is interrupted, with optional structured findings:
```json
{
"findings": [
{
"title": "<≤ 80 chars, imperative>",
"body": "<valid Markdown explaining *why* this is a problem; cite files/lines/functions>",
"confidence_score": <float 0.0-1.0>,
"priority": <int 0-3>,
"code_location": {
"absolute_file_path": "<file path>",
"line_range": {"start": <int>, "end": <int>}
}
}
],
"overall_correctness": "patch is correct" | "patch is incorrect",
"overall_explanation": "<1-3 sentence explanation justifying the overall_correctness verdict>",
"overall_confidence_score": <float 0.0-1.0>
}
```
## Questions
### Why separate out its own message history?
We want the review thread to match the training of our review models as
much as possible -- that means using a custom prompt, removing user
instructions, and starting a clean chat history.
We also want to make sure the review thread doesn't leak into the parent
thread.
### Why do this as a mode, vs. sub-agents?
1. We want review to be a synchronous task, so it's fine for now to do a
bespoke implementation.
2. We're still unclear about the final structure for sub-agents. We'd
prefer to land this quickly and then refactor into sub-agents without
rushing that implementation.
## Compact feature:
1. Stops the model when the context window become too large
2. Add a user turn, asking for the model to summarize
3. Build a bridge that contains all the previous user message + the
summary. Rendered from a template
4. Start sampling again from a clean conversation with only that bridge
`ClientRequest::NewConversation` picks up the reasoning level from the user's defaults in `config.toml`, so it should be reported in `NewConversationResponse`.
This PR changes get history op to get path. Then, forking will use a
path. This will help us have one unified codepath for resuming/forking
conversations. Will also help in having rollout history in order. It
also fixes a bug where you won't see the UI when resuming after forking.
