This PR adds the ability to favorite models for external agents—writing
to the settings in the `agent_servers` key—as well as a handful of other
improvements:
- Make the cycling keybinding `alt-enter` work for the inline assistant
as well as previous user messages
- Better organized the keybinding files removing some outdated
agent-related keybinding definitions
- Renamed the inline assistant key context to "InlineAssistant" as
"PromptEditor" is old and confusing
- Made the keybindings to rate an inline assistant response visible in
the thumbs up/down button's tooltip
- Created a unified component for the model selector tooltip given we
had 3 different places creating the same element
- Make the "Cycle Favorited Models" row in the tooltip visible only if
there is more than one favorite models
Release Notes:
- agent: External agents also now support the favoriting model feature,
which comes with a handy keybinding to cycle through the favorite list.
## Summary
This PR adds support for the MCP (Model Context Protocol)
`notifications/tools/list_changed` notification, enabling dynamic tool
discovery when MCP servers add, remove, or modify their available tools
at runtime.
## Release Notes:
- Improved: MCP tools are now automatically reloaded when a context
server sends a `tools/list_changed` notification, eliminating the need
to restart the server to discover new tools.
## Changes
- Register a notification handler for `notifications/tools/list_changed`
in `ContextServerRegistry`
- Automatically reload tools when the notification is received
- Handler is registered both on initial server startup and when a server
transitions to `Running` status
## Motivation
The MCP specification includes a `notifications/tools/list_changed`
notification to inform clients when the list of available tools has
changed. Previously, Zed's agent would only load tools once when a
context server started. This meant that:
1. If an MCP server dynamically registered new tools after
initialization, they would not be available to the agent
2. The only way to refresh tools was to restart the entire context
server
3. Tools that were removed or modified would remain in the old state
until restart
## Implementation Details
The implementation follows these steps:
1. When a context server transitions to `Running` status, register a
notification handler for `notifications/tools/list_changed`
2. The handler captures a weak reference to the `ContextServerRegistry`
entity
3. When the notification is received, spawn a task that calls
`reload_tools_for_server` with the server ID
4. The existing `reload_tools_for_server` method handles fetching the
updated tool list and notifying observers
This approach is minimal and reuses existing tool-loading
infrastructure.
## Testing
- [x] Code compiles with `./script/clippy -p agent`
- The notification handler infrastructure already exists and is tested
in the codebase
- The `reload_tools_for_server` method is already tested and working
## Benefits
- Improves developer experience by enabling hot-reloading of MCP tools
- Aligns with the MCP specification's capability negotiation system
- No breaking changes to existing functionality
- Enables more flexible and dynamic MCP server implementations
## Related Issues
This implements part of the MCP specification that was already defined
in the type system but not wired up to actually handle the
notifications.
---------
Co-authored-by: Agus Zubiaga <agus@zed.dev>
This PR solves my main pain point with Zed agent: I have a long list of
available models from different providers, and I switch between a few of
them depending on the context and the project. In particular, I use the
same models from different providers depending on whether I'm working on
a personal project or at my day job. Since I only care about a few
models (none of which are in "recommended") that are scattered all over
the list, switching between them is bothersome, even using search.
This change adds a new option in `settings.json`
(`agent.favorite_models`) and the UI to manipulate it directly from the
list of available models. When any models are marked as favorites, they
appear in a dedicated section at the very top of the list. Each model
has a small icon button that appears on hover and allows to toggle
whether it's marked as favorite.
I implemented this on the UI level (i.e. there's no first-party
knowledge about favorite models in the agent itself; in theory it could
return favorite models as a group but it would make it harder to
implement bespoke UI for the favorite models section and it also
wouldn't work for text threads which don't use the ACP infrastructure).
The feature is only enabled for the native agent but disabled for
external agents because we can't easily map their model IDs to settings
and there could be weird collisions between them.
https://github.com/user-attachments/assets/cf23afe4-3883-45cb-9906-f55de3ea2a97
Closes https://github.com/zed-industries/zed/issues/31507
Release Notes:
- Added the ability to mark language models as favorites and pin them to
the top of the list. This feature is available in the native Zed agent
(including text threads and the inline assistant), but not in external
agents via ACP.
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Fixes#43165
## Problem
MCP prompts were only available in text threads, not agent threads.
Users with MCP servers that expose prompts couldn't use them in the main
agent panel.
## Solution
Added MCP prompt support to agent threads by:
- Creating `ContextServerPromptRegistry` to track MCP prompts from
context servers
- Subscribing to context server events to reload prompts when MCP
servers start/stop
- Converting MCP prompts to available commands that appear in the slash
command menu
- Integrating prompt expansion into the agent message flow
## Testing
Tested with a custom MCP server exposing `explain-code` and
`write-tests` prompts. Prompts now appear in the `/` slash command menu
in agent threads.
Release Notes:
- Added MCP prompt support to agent threads. Prompts from MCP servers
now appear in the slash command menu when typing `/` in agent threads.
---------
Co-authored-by: Agus Zubiaga <agus@zed.dev>
Closes#26823
Release Notes:
- Added support for customising the prompt used for generating commit
message in the rules library
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
This re-introduces the `save_file` and `restore_file_from_disk` agent
tools that were reverted in #44949.
I pushed that original PR without trying it just to get the build off my
machine, but I had missed a step: the tools weren't added to the default
profile settings in `default.json`, so they were never enabled even
though the code was present.
## Changes
- Add `save_file` and `restore_file_from_disk` to the "write" profile in
`default.json`
- Add `Thread::has_tool()` method to check tool availability at runtime
- Make `edit_file_tool`'s dirty buffer error message conditional on
whether `save_file`/`restore_file_from_disk` tools are available (so the
agent gets appropriate guidance based on what tools it actually has)
- Update test to match new conditional error message behavior
Release Notes:
- Added `save_file` and `restore_file_from_disk` agent tools to handle
dirty buffers when editing files
Release Notes:
- Added `save_file` and `restore_file_from_disk` tools to the agent,
allowing it to resolve dirty buffer conflicts when editing files. When
the agent encounters a file with unsaved changes, it will now ask
whether you want to keep or discard those changes before proceeding.
Adds an optional `timeout_ms` parameter to the terminal tool that allows
bounding the runtime of shell commands. When the timeout expires, the
running terminal task is killed and the tool returns with the partial
output captured so far.
## Summary
This PR adds the ability for the agent to specify a maximum runtime when
invoking the terminal tool. This helps prevent indefinite hangs when
running commands that might wait for network, user prompts, or long
builds/tests.
## Changes
- Add `timeout_ms` field to `TerminalToolInput` schema
- Extend `TerminalHandle` trait with `kill()` method
- Implement `kill()` for `AcpTerminalHandle` and `EvalTerminalHandle`
- Race terminal exit against timeout, killing on expiry
- Update system prompt to recommend using timeouts for long-running
commands
- Add test for timeout behavior
- Update `.rules` to document GPUI executor timers for tests
## Testing
- Added `test_terminal_tool_timeout_kills_handle` which verifies that
when a timeout is specified and expires, the terminal handle is killed
and the tool returns with partial output.
- All existing agent tests pass.
Release Notes:
- agent: Added optional `timeout_ms` parameter to the terminal tool,
allowing the agent to bound command runtime and prevent indefinite hangs
We were defining these in multiple places and also weren't leveraging
the ids the agents were already providing.
This should make sure we use them consistently and avoid issues in the
future.
Release Notes:
- N/A
Uses the latest version of the SDK + schema crate. A bit painful because
we needed to move to `#[non_exhaustive]` on all of these structs/enums,
but will be much easier going forward.
Also, since we depend on unstable features, I am pinning the version so
we don't accidentally introduce compilation errors from other update
cycles.
Release Notes:
- N/A
Adds clarifying examples and information about the `include_pattern`
parameter for the grep tool, analogous to the `path` parameter of the
read and list tools and others.
The lack of clarity led to unexpected agent behavior described in
#41189. This change is confirmed to improve the precision of Claude
Sonnet 4.5 when searching for code in an empty conversation (without it,
it leaves out the root directory in the query).
Closes#41189.
```
Release Notes:
- Clarify grep tool description to improve agent precision when using it with the `include_pattern` parameter
```
It is easy for us to get the two fields out of sync causing weird
problems, there is no reason to have both here so.
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Co-authored by: Antonio Scandurra <antonio@zed.dev>
Follow up to #39021.
<img width="576" height="141" alt="image"
src="https://github.com/user-attachments/assets/c89885a4-e664-4614-9bb0-86442dff34ee"
/>
- Add migration to remove `source` tag because `ContextServerSettings`
is now untagged
- Fix typos in context server modal
- PR seems to have removed the `test_action_namespaces` test, which I
brought back in this PR
Release Notes:
- Fixed an issue where the `source` property of MCP settings would show
up as unrecognised
This PR adds back the footer with the "Configure" button in the model
selector but only when the seeing it from the Zed agent (or inline
assistant/text threads). I had removed it a while back because seeing
the "Configure" button, which takes you to the agent panel settings
view, when clicking from an external agent didn't make much sense, given
there's nothing model-wise you can configure from Zed (at least yet) for
an external agent.
This also makes the button in the footer a bit nicer by making it full
screen and displaying a keybinding, so that you can easily do the whole
"trigger model selector → go to settings view" all with the keyboard.
<img width="400" height="870" alt="Screenshot 2025-11-21 at 10 38@2x"
src="https://github.com/user-attachments/assets/c14f2acf-b793-4bc1-ac53-8a8a53b219e6"
/>
Release Notes:
- N/A
This PR simplifies error and event handling by removing the
`Ok(LanguageModelCompletionEvent::Status(CompletionRequestStatus::Failed)))`
state from the stream returned by `LanguageModel::stream_completion()`,
by changing it into an `Err(LanguageModelCompletionError)`. This was
done by collapsing the valid `CompletionRequestStatus` values into
`LanguageModelCompletionEvent`.
Release Notes:
- N/A
---------
Co-authored-by: Michael Benfield <mbenfield@zed.dev>
<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:"
Since we removed the filtering step during context gathering, we want
the model to perform more targeted searches. This PR tweaks search tool
schema allowing the model to search within syntax nodes such as `impl`
blocks or methods.
This is what the query schema looks like now:
```rust
/// Search for relevant code by path, syntax hierarchy, and content.
#[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
pub struct SearchToolQuery {
/// 1. A glob pattern to match file paths in the codebase to search in.
pub glob: String,
/// 2. Regular expressions to match syntax nodes **by their first line** and hierarchy.
///
/// Subsequent regexes match nodes within the full content of the nodes matched by the previous regexes.
///
/// Example: Searching for a `User` class
/// ["class\s+User"]
///
/// Example: Searching for a `get_full_name` method under a `User` class
/// ["class\s+User", "def\sget_full_name"]
///
/// Skip this field to match on content alone.
#[schemars(length(max = 3))]
#[serde(default)]
pub syntax_node: Vec<String>,
/// 3. An optional regular expression to match the final content that should appear in the results.
///
/// - Content will be matched within all lines of the matched syntax nodes.
/// - If syntax node regexes are provided, this field can be skipped to include as much of the node itself as possible.
/// - If no syntax node regexes are provided, the content will be matched within the entire file.
pub content: Option<String>,
}
```
We'll need to keep refining this, but the core implementation is ready.
Release Notes:
- N/A
---------
Co-authored-by: Ben <ben@zed.dev>
Co-authored-by: Max <max@zed.dev>
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
Sometimes, inside the edit agent, Sonnet thinks that it's doing a tool
call and closes its response with `</parameter></invoke>` instead of
properly closing </new_text>.
A better but more labor-intensive way of fixing this would be switching
to streaming tool calls for LLMs that support it.
Closes#39921
Release Notes:
- Fixed Sonnet's regression with inserting `</parameter></invoke>`
sometimes
Closes#41478
Release Notes:
- Fixed#41478
<img width="459" height="916" alt="Screenshot 2025-10-29 at 1 31 26 PM"
src="https://github.com/user-attachments/assets/1d5b9fdf-9800-44e4-bdd5-f0964f93625f"
/>
> caused by using haiku 4.5 from the anthropic provider and then
swapping to sonnet 3.7 through zed, doing this does mess with prompt
caching but a model swap already invalidates that so it shouldn't have
any cost impact on end users
Was previously sending all of the tools to the LLM on the first message
of a conversation regardless of the selected agent profile. This added
extra context, and tended to create scenarios where the LLM would
attempt to use the tool and it would fail since it was not available.
To reproduce, create a new conversation where you ask the Minimal mode
which tools it has access to, or try to write to a file.
Release Notes:
- Fixed an issue in the agent where all tools would be presented as
available even when using the `Minimal` profile
---------
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
