1# Agent Settings
2
3Learn about all the settings you can customize in Zed's Agent Panel.
4
5## Model Settings {#model-settings}
6
7### Default Model {#default-model}
8
9If you're using Zed's hosted LLM service, it sets `claude-sonnet-4` as the default model.
10But if you're not subscribed to the hosted service or simply just want to change it, you can do it so either via the model dropdown in the Agent Panel's bottom-right corner or by manually editing the `default_model` object in your settings:
11
12```json
13{
14 "agent": {
15 "version": "2",
16 "default_model": {
17 "provider": "zed.dev",
18 "model": "gpt-4o"
19 }
20 }
21}
22```
23
24### Feature-specific Models {#feature-specific-models}
25
26Assign distinct and specific models for the following AI-powered features in Zed:
27
28- Thread summary model: Used for generating thread summaries
29- Inline assistant model: Used for the inline assistant feature
30- Commit message model: Used for generating Git commit messages
31
32```json
33{
34 "agent": {
35 "version": "2",
36 "default_model": {
37 "provider": "zed.dev",
38 "model": "claude-sonnet-4"
39 },
40 "inline_assistant_model": {
41 "provider": "anthropic",
42 "model": "claude-3-5-sonnet"
43 },
44 "commit_message_model": {
45 "provider": "openai",
46 "model": "gpt-4o-mini"
47 },
48 "thread_summary_model": {
49 "provider": "google",
50 "model": "gemini-2.0-flash"
51 }
52 }
53}
54```
55
56> If a model isn't set for one of these features, they automatically fall back to using the default model.
57
58### Alternative Models for Inline Assists {#alternative-assists}
59
60The Inline Assist feature in particular has the capacity to perform multiple generations in parallel using different models.
61That is possible by assigning more than one model to it, taking the configuration shown above one step further.
62
63When configured, the inline assist UI will surface controls to cycle between the outputs generated by each model.
64
65The models you specify here are always used in _addition_ to your [default model](#default-model).
66
67For example, the following configuration will generate two outputs for every assist.
68One with Claude Sonnet 4 (the default model), and one with GPT-4o.
69
70```json
71{
72 "agent": {
73 "default_model": {
74 "provider": "zed.dev",
75 "model": "claude-sonnet-4"
76 },
77 "inline_alternatives": [
78 {
79 "provider": "zed.dev",
80 "model": "gpt-4o"
81 }
82 ]
83 }
84}
85```
86
87### Model Temperature
88
89Specify a custom temperature for a provider and/or model:
90
91```json
92"model_parameters": [
93 // To set parameters for all requests to OpenAI models:
94 {
95 "provider": "openai",
96 "temperature": 0.5
97 },
98 // To set parameters for all requests in general:
99 {
100 "temperature": 0
101 },
102 // To set parameters for a specific provider and model:
103 {
104 "provider": "zed.dev",
105 "model": "claude-sonnet-4",
106 "temperature": 1.0
107 }
108],
109```
110
111## Agent Panel Settings {#agent-panel-settings}
112
113Note that some of these settings are also surfaced in the Agent Panel's settings UI, which you can access either via the `agent: open configuration` action or by the dropdown menu on the top-right corner of the panel.
114
115### Default View
116
117Use the `default_view` setting to change the default view of the Agent Panel.
118You can choose between `thread` (the default) and `text_thread`:
119
120```json
121{
122 "agent": {
123 "default_view": "text_thread"
124 }
125}
126```
127
128### Auto-run Commands
129
130Control whether you want to allow the agent to run commands without asking you for permission.
131
132```json
133{
134 "agent": {
135 "always_allow_tool_actions": "true"
136 }
137}
138```
139
140> This setting is available via the Agent Panel's settings UI.
141
142### Single-file Review
143
144Control whether you want to see review actions (accept & reject) in single buffers after the agent is done performing edits.
145
146```json
147{
148 "agent": {
149 "single_file_review": "true"
150 }
151}
152```
153
154When set to false, these controls are only available in the multibuffer review tab.
155
156> This setting is available via the Agent Panel's settings UI.
157
158### Sound Notification
159
160Control whether you want to hear a notification sound when the agent is done generating changes or needs your input.
161
162```json
163{
164 "agent": {
165 "play_sound_when_agent_done": "true"
166 }
167}
168```
169
170> This setting is available via the Agent Panel's settings UI.
171
172### Modifier to Send
173
174Make a modifier (`cmd` on macOS, `ctrl` on Linux) required to send messages.
175This is encouraged for more thoughtful prompt crafting.
176
177```json
178{
179 "agent": {
180 "use_modifier_to_send": "true"
181 }
182}
183```
184
185> This setting is available via the Agent Panel's settings UI.
186
187### Edit Card
188
189Use the `expand_edit_card` setting to control whether edit cards show the full diff in the Agent Panel.
190It is set to `true` by default, but if set to false, the card's height is capped to a certain number of lines, requiring a click to be expanded.
191
192```json
193{
194 "agent": {
195 "expand_edit_card": "false"
196 }
197}
198```
199
200### Terminal Card
201
202Use the `expand_terminal_card` setting to control whether terminal cards show the command output in the Agent Panel.
203It is set to `true` by default, but if set to false, the card will be fully collapsed even while the command is running, requiring a click to be expanded.
204
205```json
206{
207 "agent": {
208 "expand_terminal_card": "false"
209 }
210}
211```
212
213### Feedback Controls
214
215Control whether you want to see the thumbs up/down buttons to give Zed feedback about the agent's performance.
216
217```json
218{
219 "agent": {
220 "enable_feedback": "false"
221 }
222}
223```