For now, just using Sonnet 4.5 and Opus 4.5 - I'll make a separate PR
for non-Anthropic models, in case they introduce new failures.
Release Notes:
- N/A
Release Notes:
- Fixed logging into Gemini CLI and Claude Code when remoting and
authenticating via CLI
Co-authored-by: Lukas Wirth <me@lukaswirth.dev>
Closes#35142
In addition to cleaning up the terminals, also stops the conversation.
Release Notes:
- Restoring a checkpoint now stops the agent conversation.
<img width="395" height="444" alt="Screenshot 2025-11-11 at 4 04 57 PM"
src="https://github.com/user-attachments/assets/8da68721-6e33-4d01-810d-4aa1e2f3402d"
/>
Discussed with @danilo-leal and we're going with the "it's checked in
both places" design!
Closes#40910
Release Notes:
- Recommended AI models now still appear in their normal category in
addition to "Recommended:"
Agents running in extensions need to have a root directory of the
extension's dir for installation and authentication, but *not* for the
conversation itself - otherwise the agent is running things like
terminal commands in the wrong dir.
Release Notes:
- Fixed Agent Server extensions having the current working directory of
the extension rather than the project
This makes it possible to do login via things like `cmd: "node", args:
["my-node-file.js", "login"]`
Also, that command will now use Zed's managed `node` instance.
Release Notes:
- ACP extensions can now run terminal login commands using relative
paths
As discussed with @benbrandt and @mikayla-maki:
* We now tell ACP clients we support the nonstandard `terminal-auth`
`_meta` field for terminal-based authentication
* In the future, we anticipate ACP itself supporting *some* form of
terminal-based authentication, but that hasn't been designed yet or gone
through the RFD process
* For now, this unblocks terminal-based auth
Release Notes:
- Added experimental terminal-based authentication to ACP support
Fixes a `--release`-only bug in feature-flagged agents where the feature
flag isn't picked up in some situations (unless there was a settings
change to go with it - due to an early return when settings didn't
change).
Release Notes:
- N/A
If we get a `cwd` from ACP (because e.g. `codex-acp` is driving the
terminal rather than our own PTY) then use that to display the `cwd` of
the terminal process.
Release Notes:
- N/A
This only affects `codex-acp` for now.
Not using the PTY in display-only terminals means they don't display the
login prompt (or spurious `%`s) at the end of terminal output
renderings.
Release Notes:
- N/A
Now we use GitHub Releases to detect when there's a new version of
codex-acp out, and we notify the user in the same way we do for the
other external agents.
This also moves `github_download.rs` out of the `languages` crate and
into `http_client`, because now we're not just using it for language
servers anymore, we're also using it for external agents.
Release Notes:
- N/A
---------
Co-authored-by: Lukas Wirth <me@lukaswirth.dev>
Codex needs (and future projects are anticipated to need as well) a
concept of display-only terminals. This refactors terminals to decouple
the PTY part from the display part, so that we can render terminal
changes based on a series of events - regardless of whether they're
being driven from a PTY inside Zed or from an outside source (e.g.
`codex-acp`).
Release Notes:
- N/A
This was added for Codex, but had undesirable consequences for Claude
Code (on Nightly, never made it to Preview). We're going to address this
in `codex-acp` instead.
Release Notes:
- N/A
<img width="598" height="311" alt="Screenshot 2025-09-11 at 9 39 12 PM"
src="https://github.com/user-attachments/assets/b526e648-37cf-4412-83a0-42037b9fc94d"
/>
This is for both ACP and the regular agent. Previously we would always
include the whole file, which can easily blow the context window on huge
files.
Release Notes:
- When `@mention`ing large files, the Agent Panel now send an outline of
the file instead of the whole thing.
If we get back slash-commands that aren't supported, tell the user that
this is the problem.
Release Notes:
- Improve error messages for unsupported ACP slash-commands
---------
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Right now if you open Zed, and we deserialize an agent that's behind a
feature flag (e.g. CC), we don't restore it because the feature flag
check hasn't happened yet at the time we're deserializing (due to auth
not having finished yet).
This is a simple fix: assume that if you had serialized it in the first
place, you must have had the feature flag enabled, so go ahead and
reopen it for you.
Release Notes:
- N/A
We have a bug in our ACP implementation where sometimes the
Accept/Reject buttons are disabled (and stay disabled even after the
thread has finished). I haven't found a complete fix for this yet, so in
the meantime I'm putting out the fire by making it so those buttons are
always enabled. That way you're never blocked, and the only consequence
of the bug is that sometimes they should be disabled but are enabled
instead.
Release Notes:
- N/A
Settings overrides (e.g. local project settings, server settings) can no
longer change `disable_ai` to `false` if it was `true`; they can only
change it to `true`. In other words, settings can only cause AI to be
*more* disabled, they can't undo the user's preference for no AI (or the
project's requirement not to use AI).
Release Notes:
- Settings overrides (such as local project settings) can now only
override `disable_ai` to become `true`; they can no longer cause
otherwise-disabled AI to become re-enabled.
---------
Co-authored-by: Assistant <assistant@anthropic.com>
Co-authored-by: David Kleingeld <git@davidsk.dev>
Now `always_allow_tool_actions` is only respected as the user's global
setting, not as an overridable project-local setting. This way, you
don't have to worry about switching into a project (or switching
branches within a project) and discovering that suddenly your tool calls
no longer require confirmation.
Release Notes:
- Removed always_allow_tool_actions from project-local settings (it is
now global-only)
Co-authored-by: David Kleingeld <git@davidsk.dev>
An unintended consequence of format on save is that we start (correctly)
informing the model that the file changed on disk every time the
formatter changes anything, which in turn can lead the model to things
like extra reads.
Until we have a solution in place to prevent this downside, we're going
back to not formatting on save by reverting
cb112a4012.
Release Notes:
- N/A
Richard Feldman
