Agent Settings

Agent Settings live in the AI section of the Settings Editor, which configures model providers, External Agents, and MCP servers. Open it with agent: open settings (also available from the top-right menu in the Agent Panel), which takes you straight to the AI page. You can also reach the same page with zed: open settings and selecting AI in the sidebar.

Within the AI page, LLM Providers, External Agents, and MCP Servers each open as their own sub-page under the General section.

For general settings mechanics, see Configuring Zed.

LLM Providers

The LLM Providers section configures model providers for Zed AI features, including Zed Agent, Inline Assistant, Git commit generation, thread summaries, and similar model-backed features.

Use this section to:

• sign in to supported subscription-backed providers
• enter provider API keys
• add OpenAI-compatible providers
• remove providers

For the model-access paths and provider-specific setup, see LLM Providers.

Feature-Specific Settings

Some Zed AI features have their own model or prompt settings in settings.json, including:

• agent.inline_assistant_model
• agent.commit_message_model
• agent.thread_summary_model
• agent.compaction_model
• agent.subagent_model
• agent.commit_message_instructions
• agent.inline_alternatives

Use agent.commit_message_instructions for instructions that apply only to generated Git commit messages:

{
  "agent": {
    "commit_message_instructions": "Use the Conventional Commits format: <type>(<scope>): <description>."
  }
}

For feature-specific model examples, see Feature-specific Models.

Automatic Compaction

Zed Agent can automatically compact long threads before they reach the selected model’s context window. Compaction summarizes earlier messages and keeps the conversation usable without starting a new thread.

Automatic compaction is enabled by default and runs when the thread reaches 90% of the model’s context window. You can change the threshold or disable automatic compaction in settings.json:

{
  "agent": {
    "auto_compact": {
      "enabled": true,
      "threshold": "90%"
    }
  }
}

The threshold value can be one of:

0 is not a valid threshold. If the threshold is invalid, Zed falls back to 90%.

You can compact a Zed Agent thread manually at any time by typing /compact in the Agent Panel message editor. For more on thread token usage and compaction behavior, see Token Usage and Compaction.

Compaction Model

By default, context compaction (both /compact and auto-compaction) uses the thread’s currently selected model. Set agent.compaction_model to use a different model:

{
  "agent": {
    "default_model": {
      "provider": "anthropic",
      "model": "claude-opus-4-6"
    },
    "compaction_model": {
      "provider": "anthropic",
      "model": "claude-sonnet-4-5"
    }
  }
}

Notes:

• The configured model should have a context window at least as large as the thread’s primary model for predictable behavior.

External Agents

The External Agents section configures ACP-integrated agents.

Use Add Agent to:

• Install from Registry
• Add Custom Agent

For setup details and support boundaries, see External Agents.

MCP Servers

The MCP Servers page configures Model Context Protocol servers connected to Zed.

Use Add Server (in the page header) to:

• Add Local Server
• Add Remote Server
• Install from Extensions

Each configured server row exposes:

• a Configure (gear) button to edit the server
• an Uninstall (trash) button to remove it
• a toggle to enable or disable it

For MCP setup, auth, server status, and agent-path boundaries, see MCP.

Related Configuration

Some AI settings are not configured on the AI settings pages:

Feature-Specific Models

Zed supports feature-specific model settings for Inline Assistant, Git commit generation, thread summaries, and subagents. Configure these in settings when you need a different model for a specific workflow.

See LLM Providers for model access, and All Settings for the complete settings reference.

Subagent Models

By default, Zed Agent’s subagents inherit the parent thread’s model, thinking setting, reasoning effort, and speed. Set agent.subagent_model in settings.json to choose separate model, reasoning, and speed settings. You can override the model by asking the agent to use a specific model for a subagent. When that model matches the provider and model in agent.subagent_model, its enable_thinking, effort, and speed settings are preserved. This setting does not configure External Agents.

Configure this setting with a provider and model. To choose a reasoning effort for a model that supports it, also set enable_thinking and effort:

{
  "agent": {
    "subagent_model": {
      "effort": "low",
      "enable_thinking": true,
      "model": "claude-opus-4-6",
      "provider": "anthropic"
    }
  }
}

Choose an effort value supported by the selected model. Omitting effort from this object does not inherit the parent’s effort. It leaves effort to the selected provider and model’s default behavior. Omitting enable_thinking defaults to false, although models that cannot disable thinking still use it.

To find supported effort values:

1. Choose the target provider and model from the model selector in a native Zed Agent thread, even if it is already selected, to save it as agent.default_model.
2. Enable thinking if it is off, then open the thinking effort menu. The “Change Thinking Effort” menu lists the model’s supported levels. If the model offers no effort choices, omit effort.
3. Choose a level and read agent.default_model.effort in settings.json for the exact value to use in agent.subagent_model.effort.

These controls change the current thread and saved default model settings, and can update a matching favorite model. Restore your previous selections afterward if needed.

The optional speed field accepts "fast" or "standard". It applies only to models that support fast mode. Omitting it from agent.subagent_model leaves speed unspecified rather than inheriting the parent’s speed.

When using matching agent.default_model settings, unsupported or omitted effort falls back to the selected model’s default effort, when available.

If the configured model is unavailable, subagents use the parent thread’s model and its thinking, effort, and speed settings.

The agent can only request a model for a subagent, not a separate effort or speed. Requesting the parent’s own model is not the same as requesting no specific model: the table’s rules for requested models apply instead of inheriting the parent’s current settings. Subagents with a requested or configured model do not follow later changes to the parent’s model, reasoning, or speed settings. Resuming a subagent keeps its existing model, reasoning, and speed settings.

Model Temperature

Most Zed AI features use the selected model’s default generation behavior. Use agent.model_parameters when you need to set a temperature for a provider, a model, or a specific provider/model pair.

{
  "agent": {
    "model_parameters": [
      {
        "provider": "anthropic",
        "model": "claude-sonnet-4-5",
        "temperature": 0.2
      }
    ]
  }
}

Zed checks matching entries from last to first. An entry can omit provider or model to apply more broadly. For provider-specific model configuration such as custom model entries, context windows, or gateway routing, see LLM Providers and the provider setup pages.

Rules, Skills, and Instructions

Reusable Rules have been replaced by Skills. Always-on Rules have moved to Instructions, including personal AGENTS.md and project instruction files.

Older builds or transitional UI may still refer to Rules. Use Skills for reusable task instructions and Instructions for always-on context.