Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Emanuele-web04/synara/llms.txt

Use this file to discover all available pages before exploring further.

Synara connects to Grok — xAI’s coding agent CLI — as a local child process using the Agent Communication Protocol (ACP). When you start a Grok thread, Synara launches the grok binary and manages the full session lifecycle: turns, approvals, context, and checkpoints. Grok handles reasoning and code generation using your xAI account. The Grok CLI must be installed and authorized on your machine before Synara can run Grok sessions.

​
Prerequisites

  • Grok CLI installed — follow the xAI Grok CLI setup instructions to install it.
  • Grok CLI authorized with a valid xAI account (run grok in your terminal and complete the login flow).
Synara does not install or manage the Grok CLI. If the grok binary is not found on your PATH, set binaryPath to the correct location in settings.json and restart Synara.

​
Configuration

Grok is configured under providers.grok in settings.json. All fields are optional. 
{
  "providers": {
    "grok": {
      "enabled": true,
      "binaryPath": "grok",
      "customModels": []
    }
  }
}
​
enabled
boolean
default:"true"
Set to false to disable the Grok provider entirely. Synara will skip it during startup discovery and will not allow new Grok threads.
​
binaryPath
string
default:"grok"
The command or full path used to launch the Grok CLI. Defaults to "grok", resolved via your PATH. Use an absolute path if your installation is not on PATH — for example, "/usr/local/bin/grok".
​
customModels
string[]
default:"[]"
Additional model slugs to display in the model picker alongside Synara’s built-in Grok model list. Useful for preview or beta Grok models.

​
Model options

Grok model options let you control how much reasoning the model performs before replying. 
// From packages/contracts/src/model.ts
export const GROK_REASONING_EFFORT_OPTIONS = ["none", "low", "medium", "high"] as const;

export const GrokModelOptions = Schema.Struct({
  reasoningEffort: Schema.optional(Schema.Literals(GROK_REASONING_EFFORT_OPTIONS)),
});
​
reasoningEffort
string
Controls the reasoning effort the Grok model applies before generating a response.
ValueDescription
noneNo reasoning — immediate, direct responses.
lowMinimal reasoning — fast responses with light analysis. Default for Grok Build models.
mediumModerate reasoning — balanced speed and thoroughness.
highMaximum reasoning effort — slower but most thorough.

​
Built-in model roster

SlugNameDefault effort
grok-buildGrok 4.3low
grok-build-0.1Grok Build 0.1low
The default model for new Grok threads is grok-build (Grok 4.3).

​
Related

Build docs developers (and LLMs) love

Get started for freeTalk to us