Struct Config 

pub struct Config {

    pub project: String,
    pub next_id: u32,
    pub auto_close_parent: bool,
    pub run: Option<String>,
    pub plan: Option<String>,
    pub max_loops: u32,
    pub max_concurrent: u32,
    pub poll_interval: u32,
    pub extends: Vec<String>,
    pub rules_file: Option<String>,
    pub file_locking: bool,
    pub worktree: bool,
    pub on_close: Option<String>,
    pub on_fail: Option<String>,
    pub post_plan: Option<String>,
    pub verify_timeout: Option<u64>,
    pub review: Option<ReviewConfig>,
    pub user: Option<String>,
    pub user_email: Option<String>,
    pub auto_commit: bool,
    pub commit_template: Option<String>,
    pub research: Option<String>,
    pub run_model: Option<String>,
    pub plan_model: Option<String>,
    pub review_model: Option<String>,
    pub research_model: Option<String>,
    pub batch_verify: bool,
    pub memory_reserve_mb: u64,
    pub notify: Option<NotifyConfig>,
}

Project-level mana configuration, loaded from .mana/config.yaml.

All fields have sensible defaults; only project and next_id are required in the YAML file. Optional fields are omitted when serialized to keep config files minimal.

Use Config::load_with_extends to load with inherited values from parent configs listed in the extends field.

Fields

project: String

next_id: u32

auto_close_parent: bool

Auto-close parent units when all children are closed/archived (default: true)

run: Option<String>

Shell command template for --run. Use {id} as placeholder for unit ID. Example: claude -p "implement unit {id} and run mana close {id}". If unset, --run will print an error asking the user to configure it.

plan: Option<String>

Shell command template for planning large units. Uses {id} placeholder. If unset, plan operations will print an error asking the user to configure it.

max_loops: u32

Maximum agent loops before stopping (default: 10, 0 = unlimited)

max_concurrent: u32

Maximum parallel agents for mana run (default: 4)

poll_interval: u32

Seconds between polls in –watch mode (default: 30)

extends: Vec<String>

Paths to parent config files to inherit from (lowest to highest priority). Supports ~/ for home directory. Paths are relative to the project root.

rules_file: Option<String>

Path to project rules file, relative to .mana/ directory (default: “RULES.md”). Contents are injected into every mana context output.

file_locking: bool

Enable file locking for concurrent agents (default: false). When enabled, agents lock files listed in unit paths on spawn and lock-on-write during execution. Prevents concurrent agents from clobbering the same file.

worktree: bool

Enable git worktree isolation for parallel agents (default: false). When enabled, mana run creates a separate git worktree for each agent. Each agent works in its own directory, preventing file contention. On mana close, the worktree branch is merged back to main.

on_close: Option<String>

Shell command template to run after a unit is successfully closed. Supports template variables: {id}, {title}, {status}, {branch}. Runs asynchronously — failures are logged but don’t affect the close.

on_fail: Option<String>

Shell command template to run after a verify attempt fails. Supports template variables: {id}, {title}, {attempt}, {output}, {branch}. Runs asynchronously — failures are logged but don’t affect the operation.

post_plan: Option<String>

Shell command template to run after mana plan creates children. Supports template variables: {id}, {parent}, {children}, {branch}. Runs asynchronously — failures are logged but don’t affect the plan.

verify_timeout: Option<u64>

Default timeout in seconds for verify commands (default: None = no limit). Per-unit verify_timeout overrides this value.

review: Option<ReviewConfig>

Adversarial review configuration (mana review / mana run --review). Optional — review is disabled if not configured.

user: Option<String>

User identity name (e.g., “alice”). Used for claimed_by and created_by.

user_email: Option<String>

User email (e.g., “alice@co”). Optional, for git integration.

auto_commit: bool

Automatically commit all changes when a unit is closed (default: false). Creates a commit with message based on commit_template. Skipped in worktree mode (worktree already commits).

commit_template: Option<String>

Template for auto-commit messages. Placeholders: {id}, {title}, {parent_id}, {labels}. Default: “feat(unit-{id}): {title}”

Keep {id} in the template so mana diff <id> can find the unit’s commit.

research: Option<String>

Shell command template for project-level research (mana plan with no ID). Uses {parent_id} as placeholder for the parent unit that groups findings. Falls back to plan template with a research-oriented prompt if unset.

run_model: Option<String>

Model to use for implementing units (mana run). Substituted into {model} in templates.

plan_model: Option<String>

Model to use for planning/splitting (mana plan). Substituted into {model} in templates.

review_model: Option<String>

Model to use for adversarial review (mana run --review). Substituted into {model} in templates.

research_model: Option<String>

Model to use for project research (mana plan with no args). Substituted into {model} in templates.

batch_verify: bool

Enable runner-mediated batch verification (default: false). When enabled, mana run sets MANA_BATCH_VERIFY=1 on spawned agents. Agents signal completion without running verify inline; the runner collects AwaitingVerify units and runs each unique verify command once.

memory_reserve_mb: u64

Minimum available system memory (MB) to keep free when spawning agents. When set to a non-zero value, mana run checks available system memory before each agent spawn. If available memory is below this threshold, dispatch pauses until a running agent finishes and frees memory. Default: 0 (disabled). Recommended: 2048–4096 on a 16GB machine.

notify: Option<NotifyConfig>

Notification settings for human-facing alerts (push notifications, desktop alerts, webhook pings). Separate from on_close/on_fail workflow hooks — those are for automation, these are for humans.

Implementations

impl Config

pub fn load(mana_dir: &Path) -> Result<Config, Error>

Load config from .mana/config.yaml inside the given units directory.

pub fn load_with_extends(mana_dir: &Path) -> Result<Config, Error>

Load config with inheritance from extended configs.

Resolves the extends field, loading parent configs and merging inheritable fields. Local values take precedence over extended values. Fields project, next_id, and extends are never inherited.

pub fn save(&self, mana_dir: &Path) -> Result<(), Error>

Save config to .mana/config.yaml inside the given units directory.

pub fn rules_path(&self, mana_dir: &Path) -> PathBuf

Return the path to the project rules file. Defaults to .mana/RULES.md if rules_file is not set. The path is resolved relative to the units directory.

pub fn increment_id(&mut self) -> u32

Return the current next_id and increment it for the next call.

Trait Implementations

impl Debug for Config

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

Formats the value using the given formatter.

impl Default for Config

fn default() -> Config

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for Config

fn deserialize<__D>( __deserializer: __D, ) -> Result<Config, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for Config

fn eq(&self, other: &Config) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for Config

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for Config