We get spurrious reports that the model writes fenced code blocks
without an info tag which then causes auto-language detection in the
extension to incorrectly highlight the code and show the wrong language.
The model should really always include a tag when it can.
This PR adds support for streamable HTTP MCP servers when the
`experimental_use_rmcp_client` is enabled.
To set one up, simply add a new mcp server config with the url:
```
[mcp_servers.figma]
url = "http://127.0.0.1:3845/mcp"
```
It also supports an optional `bearer_token` which will be provided in an
authorization header. The full oauth flow is not supported yet.
The config parsing will throw if it detects that the user mixed and
matched config fields (like command + bearer token or url + env).
The best way to review it is to review `core/src` and then
`rmcp-client/src/rmcp_client.rs` first. The rest is tests and
propagating the `Transport` struct around the codebase.
Example with the Figma MCP:
<img width="5084" height="1614" alt="CleanShot 2025-09-26 at 13 35 40"
src="https://github.com/user-attachments/assets/eaf2771e-df3e-4300-816b-184d7dec5a28"
/>
The [official Rust
SDK](https://github.com/modelcontextprotocol/rust-sdk/tree/57fc428c578a1a3fe851ee0838bf068bda120eb3)
has come a long way since we first started our mcp client implementation
5 months ago and, today, it is much more complete than our own
stdio-only implementation.
This PR introduces a new config flag `experimental_use_rmcp_client`
which will use a new mcp client powered by the sdk instead of our own.
To keep this PR simple, I've only implemented the same stdio MCP
functionality that we had but will expand on it with future PRs.
---------
Co-authored-by: pakrym-oai <pakrym@openai.com>
# External (non-OpenAI) Pull Request Requirements
Currently, mcp server fail to start with:
```
🖐 MCP client for `<CLIENT>` failed to start: missing field `user_agent`
````
It isn't clear to me yet why this is happening. My understanding is that
this struct is simply added as a new field to the response but this
should fix it until I figure out the full story here.
<img width="714" height="262" alt="CleanShot 2025-09-10 at 13 58 59"
src="https://github.com/user-attachments/assets/946b1313-5c1c-43d3-8ae8-ecc3de3406fc"
/>
The previous config approach had a few issues:
1. It is part of the config but not designed to be used externally
2. It had to be wired through many places (look at the +/- on this PR
3. It wasn't guaranteed to be set consistently everywhere because we
don't have a super well defined way that configs stack. For example, the
extension would configure during newConversation but anything that
happened outside of that (like login) wouldn't get it.
This env var approach is cleaner and also creates one less thing we have
to deal with when coming up with a better holistic story around configs.
One downside is that I removed the unit test testing for the override
because I don't want to deal with setting the global env or spawning
child processes and figuring out how to introspect their originator
header. The new code is sufficiently simple and I tested it e2e that I
feel as if this is still worth it.
This PR does multiple things that are necessary for conversation resume
to work from the extension. I wanted to make sure everything worked so
these changes wound up in one PR:
1. Generate more ts types
2. Resume rollout history files rather than create a new one every time
it is resumed so you don't see a duplicate conversation in history for
every resume. Chatted with @aibrahim-oai to verify this
3. Return conversation_id in conversation summaries
4. [Cleanup] Use serde and strong types for a lot of the rollout file
parsing
We're trying to migrate from `session_id: Uuid` to `conversation_id:
ConversationId`. Not only does this give us more type safety but it
unifies our terminology across Codex and with the implementation of
session resuming, a conversation (which can span multiple sessions) is
more appropriate.
I started this impl on https://github.com/openai/codex/pull/3219 as part
of getting resume working in the extension but it's big enough that it
should be broken out.
This updates the ctrl + c behavior to clear the current prompt if there
is text and you press ctrl + c.
I also updated the ctrl + c hint text to show `^c to interrupt` instead
of `^c to quit` if there is an active conversation.
Two things I don't love:
1. You can currently interrupt a conversation with escape or ctrl + c
(not related to this PR and maybe fine)
2. The bottom row hint text always says `^c to quit` but this PR doesn't
really make that worse.
https://github.com/user-attachments/assets/6eddadec-0d84-4fa7-abcb-d6f5a04e5748
Fixes https://github.com/openai/codex/issues/3126
## Summary
- read the shell exec approval request's actual id instead of assuming
it is always 0
- use that id when validating and responding in the test
## Testing
- `cargo test -p codex-mcp-server
test_shell_command_approval_triggers_elicitation`
------
https://chatgpt.com/codex/tasks/task_i_68a6ab9c732c832c81522cbf11812be0
Right now, every time an exec ends, we emit it to history which makes it
immutable. In order to be able to update or merge successive tool calls
(which will be useful after https://github.com/openai/codex/pull/2095),
we need to retain it as the active cell.
This also changes the cell to contain the metadata necessary to render
it so it can be updated rather than baking in the final text lines when
the cell is created.
Part 1: https://github.com/openai/codex/pull/2095
Part 3: https://github.com/openai/codex/pull/2110
# Note for reviewers
The bulk of this PR is in in the new file, `parse_command.rs`. This file
is designed to be written TDD and implemented with Codex. Do not worry
about reviewing the code, just review the unit tests (if you want). If
any cases are missing, we'll add more tests and have Codex fix them.
I think the best approach will be to land and iterate. I have some
follow-ups I want to do after this lands. The next PR after this will
let us merge (and dedupe) multiple sequential cells of the same such as
multiple read commands. The deduping will also be important because the
model often reads the same file multiple times in a row in chunks
===
This PR formats common commands like reading, formatting, testing, etc
more nicely:
It tries to extract things like file names, tests and falls back to the
cmd if it doesn't. It also only shows stdout/err if the command failed.
<img width="770" height="238" alt="CleanShot 2025-08-09 at 16 05 15"
src="https://github.com/user-attachments/assets/0ead179a-8910-486b-aa3d-7d26264d751e"
/>
<img width="348" height="158" alt="CleanShot 2025-08-09 at 16 05 32"
src="https://github.com/user-attachments/assets/4302681b-5e87-4ff3-85b4-0252c6c485a9"
/>
<img width="834" height="324" alt="CleanShot 2025-08-09 at 16 05 56 2"
src="https://github.com/user-attachments/assets/09fb3517-7bd6-40f6-a126-4172106b700f"
/>
Part 2: https://github.com/openai/codex/pull/2097
Part 3: https://github.com/openai/codex/pull/2110
## Summary
- support `codex logout` via new subcommand and helper that removes the
stored `auth.json`
- expose a `logout` function in `codex-login` and test it
- add `/logout` slash command in the TUI; command list is filtered when
not logged in and the handler deletes `auth.json` then exits
## Testing
- `just fix` *(fails: failed to get `diffy` from crates.io)*
- `cargo test --all-features` *(fails: failed to get `diffy` from
crates.io)*
------
https://chatgpt.com/codex/tasks/task_i_68945c3facac832ca83d48499716fb51
This sets up the scaffolding and basic flow for a TUI onboarding
experience. It covers sign in with ChatGPT, env auth, as well as some
safety guidance.
Next up:
1. Replace the git warning screen
2. Use this to configure default approval/sandbox modes
Note the shimmer flashes are from me slicing the video, not jank.
https://github.com/user-attachments/assets/0fbe3479-fdde-41f3-87fb-a7a83ab895b8
This lets us show an accumulating diff across all patches in a turn.
Refer to the docs for TurnDiffTracker for implementation details.
There are multiple ways this could have been done and this felt like the
right tradeoff between reliability and completeness:
*Pros*
* It will pick up all changes to files that the model touched including
if they prettier or another command that updates them.
* It will not pick up changes made by the user or other agents to files
it didn't modify.
*Cons*
* It will pick up changes that the user made to a file that the model
also touched
* It will not pick up changes to codegen or files that were not modified
with apply_patch
This adds a tool the model can call to update a plan. The tool doesn't
actually _do_ anything but it gives clients a chance to read and render
the structured plan. We will likely iterate on the prompt and tools
exposed for planning over time.
1. Emit call_id to exec approval elicitations for mcp client convenience
2. Remove the `-retry` from the call id for the same reason as above but
upstream the reset behavior to the mcp client
1. Added an elicitation for `approve-patch` which is very similar to
`approve-exec`.
2. Extracted both elicitations to their own files to prevent
`codex_tool_runner` from blowing up in size.
In order to to this, I created a new `chatgpt` crate where we can put
any code that interacts directly with ChatGPT as opposed to the OpenAI
API. I added a disclaimer to the README for it that it should primarily
be modified by OpenAI employees.
https://github.com/user-attachments/assets/bb978e33-d2c9-4d8e-af28-c8c25b1988e8
Gabriel PealandGitHub
