Struct AgentConfig 

pub struct AgentConfig {

    pub name: String,
    pub description: Option<String>,
    pub system_prompt: String,
    pub allowed_tools: Vec<String>,
    pub disallowed_tools: Vec<String>,
    pub model: Option<String>,
    pub base_url: Option<String>,
    pub provider: Option<String>,
    pub max_tokens: Option<u32>,
    pub temperature: Option<f64>,
    pub thinking_budget: Option<u32>,
    pub reasoning_effort: Option<String>,
    pub max_context_tokens: Option<usize>,
    pub max_iterations: Option<u32>,
    pub write_access: bool,
    pub skip_memory: bool,
}

Top-level agent configuration loaded from JSON.

Place agent JSON files in:

• agents/ (project-level, highest priority)
• ~/.config/koda/agents/ (user-level)
• Built-in (embedded at compile time, lowest priority)

Example

{
  "name": "testgen",
  "system_prompt": "You are a test generation specialist.",
  "model": "gemini-2.5-flash",
  "write_access": true,
  "allowed_tools": ["Read", "Write", "Edit", "Bash", "Grep", "Glob"],
  "max_iterations": 20
}

Fields

• name — Agent identifier (used with InvokeAgent)
• system_prompt — Behavioral instructions for the LLM
• allowed_tools — Allowlist (empty = all tools available)
• disallowed_tools — Denylist (excluded even if allowed_tools is empty)
• model — Override the default model (e.g. "gemini-2.5-flash" for cheap workers)
• write_access — Grant Write/Edit/Delete tools (default: false)
• max_iterations — Cap inference loops (prevents runaway agents)

Fields

name: String

Agent identifier.

description: Option<String>

One-line description shown in the main agent’s system prompt listing. Optional — agents without a description are listed by name only.

system_prompt: String

System prompt template.

allowed_tools: Vec<String>

Allowlisted tool names (empty = all tools).

disallowed_tools: Vec<String>

Denylisted tool names — excluded even if allowed_tools is empty.

model: Option<String>

Override model identifier.

base_url: Option<String>

Override API base URL.

provider: Option<String>

Override provider type.

max_tokens: Option<u32>

Override max output tokens.

temperature: Option<f64>

Override temperature.

thinking_budget: Option<u32>

Override thinking budget (Anthropic extended thinking).

reasoning_effort: Option<String>

Override reasoning effort (OpenAI reasoning models).

max_context_tokens: Option<usize>

Override max context window tokens.

max_iterations: Option<u32>

Override max inference iterations.

write_access: bool

Grant write access (Write/Edit/Delete tools). Default: false. Sub-agents are read-only by default (principle of least privilege). Set to true for agents that need to create or modify files.

skip_memory: bool

Skip injecting project/global memory into the system prompt. Default: false. Read-only agents (explore, plan) don’t need memory context — skipping it saves tokens without affecting their ability to search the codebase.

Trait Implementations

impl Clone for AgentConfig

fn clone(&self) -> AgentConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AgentConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for AgentConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.