Configuration Reference

Complete reference for all Automergent configuration options, environment variables, and CLI flags.

Config File Locations

Automergent uses a layered configuration system. Settings are loaded in order, with later sources overriding earlier ones:

Configuration Layer Priority

Hard-coded defaults — built into the binary
Global config — ~/.automergent/config.yaml
Project config — .automergent/config.yaml (overrides global)
Environment variables — take precedence over file-based config
CLI flags — highest precedence, override everything

Layer	Path	Description
Global	`~/.automergent/config.yaml`	User-wide settings applied to all projects. Created on first run.
Project	`.automergent/config.yaml`	Project-specific overrides. Committed to version control for team-wide settings.
Environment	Shell environment	`AUTOMERGENT_*` prefixed variables override file-based config. API keys are always env-first.
CLI Flags	Command line	One-shot overrides. Not persisted. Highest priority.

Note: The config file supports both YAML and JSON formats. YAML is the default and recommended format. You can specify an alternate config file with --config <path>.

Full Config Schema

Below is every configuration option organized by section. Each option shows its type, default value, and constraints.

Core

Key	Type	Default	Description
`provider` required	`string`	`"google"`	AI provider. One of: `google-aistudio`, `google-vertex`, `custom`, or legacy aliases `google`, `openai`, `anthropic`, `deepseek`, `ollama`.
`model` required	`string`	`"gemini-3.6-flash"`	Model identifier. Must be non-empty.
`mode` required	`string`	`"edit"`	Operating mode. One of: `edit`, `code`, `chat`, `agent`.
`theme`	`string`	`"modern"`	UI theme. One of: `modern`, `catppuccin`, `dracula`, `nord`, `gruvbox`, `onedark`, `monokai`, `tokyonight`, `solarized-dark`, `solarized-light`, `high-contrast`.
`keybindings`	`string`	`"default"`	Keybinding preset. One of: `default`, `vim`, `emacs`.
`layout`	`string`	`"default"`	UI layout mode. One of: `default`, `minimal`, `full`, `split`.

Session

Key	Type	Default	Description
`autoSave`	`bool`	`true`	Automatically save session state at checkpoint intervals.
`checkpointInterval`	`int`	`5`	Interval in minutes between auto-checkpoints. Minimum: 1.
`sessionDir`	`string`	`~/.automergent/sessions`	Directory where session transcripts are stored.
`maxSessions`	`int`	`100`	Maximum number of sessions to retain. Oldest are pruned first.
`maxSessionAge`	`duration`	unlimited	Maximum age of sessions (e.g., `"30d"`, `"72h"`). Empty means no age limit.

Context Management

Tip: Automergent's context system manages a 1M-token ceiling. The maxContextTokens setting controls the soft limit per session; auto-compression kicks in when usage exceeds the configured thresholds.

Key	Type	Default	Description
`maxContextTokens`	`int`	`128000`	Soft context window limit. Hard ceiling is 1,048,576 (1M). Larger values are clamped.
`maxToolOutputChars`	`int`	`32000`	Maximum characters retained from tool output before truncation.
`warnAtContextFraction`	`float`	`0.8`	Warn when context usage exceeds this fraction (0–1) of `maxContextTokens`.
`autoCompressAt`	`float`	`0.9`	Automatically compress context when usage exceeds this fraction (0–1).
`compressionKeepRecent`	`int`	`10`	Number of most recent messages to preserve during auto-compression.

Model

Key	Type	Default	Description
`fastModel`	`string`	empty (use main model)	Cheaper model for read-only subagents (explore, review). Empty falls back to the primary model.
`thinkingBudget`	`int`	provider default	Override Gemini thinking token budget when > 0.
`effort`	`string`	provider default	Thinking effort level. One of: `minimal`, `low`, `medium`, `high`, `max`.

File Handling

Key	Type	Default	Description
`maxAutoReadFileSize`	`int`	`524288` (512 KB)	Maximum file size in bytes for automatic reads. Larger files require explicit `read`.
`maxTreeFiles`	`int`	`1000`	Maximum files shown in directory tree output.
`maxTreeDepth`	`int`	`10`	Maximum directory depth in tree output.
`excludePatterns`	`[]string`	see below	Glob patterns to exclude from tree/glob/search operations.
`contextFiles`	`[]string`	empty	Extra files to inject into context at session start (paths relative to project root).
`skillsDir`	`string`	`~/.automergent/skills`	Directory containing custom skill definitions.

Default exclude patterns:

# Applied automatically unless overridden
excludePatterns:
  - ".git"
  - "node_modules"
  - ".venv"
  - "__pycache__"
  - "*.pyc"
  - "*.o"
  - "*.a"
  - "vendor"

Display

Key	Type	Default	Description
`noAnimation`	`bool`	`false`	Disable TUI animations (useful for CI or slow terminals).
`noColor`	`bool`	`false`	Disable ANSI color output entirely.
`noTui`	`bool`	`false`	Run in non-interactive CLI-only mode. Never persisted to config file.
`output`	`string`	`"text"`	Output format for non-TUI mode.
`quiet`	`bool`	`false`	Suppress non-essential output.
`verbose`	`bool`	`false`	Enable verbose/detailed output.

Edit

Key	Type	Default	Description
`editReview`	`bool`	`false`	Route file mutations through pending-review flow. Proposals render in the diff pane and touch disk only on accept.

Security

Warning: Running Automergent as root or without a sandbox increases risk. Always use requireGitForAutoModes: true in production environments.

Key	Type	Default	Description
`security.sandbox`	`string`	`"auto"`	Sandbox mode. One of: `off`, `auto`, `strict`.
`security.blockedWritePaths`	`[]string`	see below	Glob patterns for paths that cannot be written to.
`security.allowedWritePaths`	`[]string`	empty	Explicit allowlist of writable paths. When non-empty, only these paths may be modified.
`security.stripEnvVarPatterns`	`[]string`	see below	Environment variable patterns to strip from tool output (prevents secret leakage).
`security.requireGitForAutoModes`	`bool`	`true`	Require a clean git working directory before enabling auto/accept-edits modes.
`security.rootRiskAcknowledged`	`bool`	`false`	Acknowledge risk of running as root. Required to disable sandbox warnings.

Default blocked write paths:

security:
  blockedWritePaths:
    - ".git/**"
    - "~/.ssh/**"
    - "~/.gnupg/**"
    - "~/.aws/**"
    - "~/.kube/**"

Default env var strip patterns:

security:
  stripEnvVarPatterns:
    - "*_SECRET"
    - "*_PASSWORD"
    - "*_TOKEN"
    - "*_KEY"
    - "AWS_*"
    - "GEMINI_*"
    - "HTTP_PROXY"
    - "HTTPS_PROXY"
    - "ALL_PROXY"
    - "NO_PROXY"
    - "http_proxy"
    - "https_proxy"
    - "all_proxy"
    - "no_proxy"

Per-Tool Configuration

Individual tools can be enabled/disabled, given custom timeouts, and configured for confirmation behavior.

Key	Type	Default	Description
`tools.<name>.enabled`	`bool`	`true`	Enable or disable the tool entirely.
`tools.<name>.confirmationRequired`	`string`	`"always"`	Confirmation policy. One of: `always`, `write`, `never`.
`tools.<name>.timeout`	`duration`	`"30s"`	Maximum execution time (e.g., `"5s"`, `"2m"`).
`tools.<name>.maxOutputBytes`	`int`	varies by tool	Maximum output bytes before truncation.

Built-in tool defaults:

tools:
  shell:
    enabled: true
    confirmationRequired: "always"
    timeout: "30s"
    maxOutputBytes: 1048576  # 1 MB
  filesystem:
    enabled: true
    confirmationRequired: "write"
    maxOutputBytes: 524288   # 512 KB
  git:
    enabled: true
    confirmationRequired: "never"
  web:
    enabled: true
    confirmationRequired: "never"
    timeout: "15s"

LSP Integration

Key	Type	Default	Description
`lsp.enabled`	`bool`	`true`	Enable Language Server Protocol integration.
`lsp.servers`	`map[string]string`	empty	Map of language → server command (e.g., `{"go": "gopls"}"`).
`lsp.startupTimeout`	`duration`	`"10s"`	Timeout for LSP server initialization.
`lsp.requestTimeout`	`duration`	`"5s"`	Timeout for individual LSP requests.

MCP Servers

Model Context Protocol (MCP) servers extend Automergent with custom tool providers.

Key	Type	Description
`mcp.servers.<name>.type`	`string`	Server type: `"stdio"` or `"sse"`.
`mcp.servers.<name>.command`	`[]string`	Command + args for stdio servers (e.g., `["npx", "-y", "@modelcontextprotocol/server-filesystem"]`).
`mcp.servers.<name>.url`	`string`	URL for SSE servers.
`mcp.servers.<name>.env`	`map[string]string`	Environment variables passed to the server process.
`mcp.servers.<name>.auth`	`string`	Authentication token for SSE servers.
`mcp.servers.<name>.timeout`	`duration`	Connection timeout (e.g., `"30s"`).

mcp:
  servers:
    filesystem:
      type: "stdio"
      command:
        - "npx"
        - "-y"
        - "@modelcontextprotocol/server-filesystem"
        - "/path/to/allowed/dir"
    remote-tools:
      type: "sse"
      url: "https://mcp.example.com/sse"
      auth: "Bearer ${MCP_TOKEN}"
      timeout: "30s"

Logging

Key	Type	Default	Description
`log.level`	`string`	`"warn"`	Log verbosity. One of: `debug`, `info`, `warn`, `error`.
`log.file`	`string`	`~/.automergent/automergent.log`	Path to the log file.
`log.maxSize`	`size`	`"50MB"`	Maximum log file size before rotation.
`log.maxBackups`	`int`	`3`	Maximum number of rotated log files to keep.

Providers

Per-provider configuration. Keys are provider identifiers (e.g., google-aistudio, openai, anthropic).

Key	Type	Description
`providers.<name>.apiKey`	`string`	API key. Use `"${ENV_VAR}"` syntax for env var interpolation.
`providers.<name>.baseUrl`	`string`	Custom API endpoint URL.
`providers.<name>.modelApiUrl`	`string`	Endpoint path for fetching model lists (e.g., `"/v1/models"`).
`providers.<name>.defaultModel`	`string`	Default model when none is configured.
`providers.<name>.orgId`	`string`	Organization ID (OpenAI).
`providers.<name>.project`	`string`	GCP project ID (Vertex AI).
`providers.<name>.location`	`string`	GCP region (Vertex AI, e.g., `"us-central1"`).
`providers.<name>.apiType`	`string`	API protocol: `"openai"`, `"anthropic"`, `"gemini"`, `"custom"`.
`providers.<name>.backend`	`string`	Provider backend: `"aistudio"` or `"vertex"` (Google only).
`providers.<name>.effort`	`string`	Default effort level for this provider.
`providers.<name>.thinkingLevel`	`string`	Default thinking level for this provider.
`providers.<name>.temperature`	`*float`	Sampling temperature (0.0–2.0). `null` uses provider default.
`providers.<name>.maxTokens`	`int`	Maximum tokens in completion. 0 = provider default.
`providers.<name>.timeoutSeconds`	`int`	Request timeout in seconds. 0 = provider default.
`providers.<name>.maxRetries`	`int`	Maximum retry attempts on failure.
`providers.<name>.headers`	`map[string]string`	Custom HTTP headers sent with requests.
`providers.<name>.models`	`map[string]ModelConfig`	Per-model overrides (see below).

Model-level config (under providers.<name>.models.<model>):

Key	Type	Description
`apiKey`	`string`	Model-specific API key override.
`baseUrl`	`string`	Model-specific endpoint override.
`displayName`	`string`	Human-readable name for UI display.
`contextLimit`	`int`	Context window size in tokens.
`inputPrice`	`float`	Price per million input tokens.
`outputPrice`	`float`	Price per million output tokens.

Git

Key	Type	Default	Description
`git.coAuthor`	`string`	`"ask"`	Co-author policy: `"always"`, `"never"`, or `"ask"`.
`git.coAuthorTrailer`	`string`	`"Co-authored-by: Automergent <automergent-bot@users.noreply.github.com>"`	Custom co-author trailer line appended to commits.

Diagnostics

Key	Type	Default	Description
`diagnostics.enabled`	`bool`	`true`	Enable compiler/diagnostic error detection.
`diagnostics.showInRead`	`bool`	`true`	Display diagnostics inline when reading files.
`diagnostics.blockOnError`	`bool`	`true`	Block agent execution when a diagnostic error is detected.
`diagnostics.blockOnWarning`	`bool`	`false`	Block agent execution when a diagnostic warning is detected.
`diagnostics.maxFileSizeBytes`	`int64`	`1048576` (1 MB)	Maximum file size to parse for diagnostics.
`diagnostics.cacheDurationSec`	`int`	`30`	Duration in seconds to cache diagnostic results.

Cache

Key	Type	Default	Description
`cache.prompt.enabled`	`bool`	`true`	Enable provider prompt caching (saves tokens on repeated prompts).

Privacy

Key	Type	Default	Description
`zeroDataRetention`	`bool`	`false`	Request zero data retention from supported providers.
`telemetry`	`bool`	`true`	Enable anonymous usage telemetry.

Debug

Key	Type	Default	Description
`debug.enabled`	`bool`	`true`	Enable debug mode for API request/response logging.
`debug.directory`	`string`	`~/.automergent/debug`	Directory to store debug logs.
`debug.saveRequests`	`bool`	`true`	Save raw API request payloads.
`debug.saveResponses`	`bool`	`true`	Save raw API response payloads.
`debug.maxFileSize`	`int`	`10`	Maximum debug log file size in MB. 0 = unlimited.

Provider Fallback

Define fallback providers that are tried when the primary provider fails. Falls back in list order.

Key	Type	Description
`providerFallback`	`[]FallbackProvider`	Ordered list of fallback providers.
`providerFallback[].provider`	`string`	Fallback provider identifier.
`providerFallback[].model`	`string`	Fallback model identifier.

providerFallback:
  - provider: "google-aistudio"
    model: "gemini-3.6-flash"
  - provider: "custom"
    model: "deepseek-chat"

Notifications

Key	Type	Default	Description
`notifications.desktop`	`bool`	`false`	Enable desktop notifications for task completion.
`notifications.bell`	`bool`	`true`	Ring terminal bell on notification.
`notifications.contextWarning`	`bool`	`true`	Show warning when context window is nearly full.

Hooks

Lifecycle hooks run shell commands at specific points in the tool execution pipeline. Hooks receive context via environment variables: AUTOMERGENT_TOOL_NAME, AUTOMERGENT_TOOL_ARGS, AUTOMERGENT_TOOL_RESULT, AUTOMERGENT_PROJECT_DIR.

Key	Type	Description
`hooks.preToolUse`	`[]Hook`	Commands executed before tool execution. Non-zero exit vetoes the tool call (stderr becomes the reason).
`hooks.postToolUse`	`[]Hook`	Commands executed after tool execution. Failures are logged, never fatal.
`hooks.*[].name`	`string`	Human-readable hook name (for logs).
`hooks.*[].command`	`string`	Shell command to execute.
`hooks.*[].timeout`	`duration`	Hook execution timeout (default: `"10s"`).

hooks:
  preToolUse:
    - name: "lint-check"
      command: "npm run lint --quiet"
      timeout: "30s"
  postToolUse:
    - name: "notify-done"
      command: "echo 'Tool completed' | pbcopy"

Timeouts

Key	Type	Default	Description
`confirmationTimeout`	`duration`	`"10m"`	Default timeout for user confirmation dialogs (e.g., tool execution approval).

Environment Variables

Environment variables override file-based configuration. API keys are always resolved from environment variables first.

Variable	Used For	Description
`GOOGLE_API_KEY`	Google AI Studio	API key for Gemini models via AI Studio. Higher priority than `GEMINI_API_KEY`.
`GEMINI_API_KEY`	Google AI Studio	Alias for Google API key. Used as fallback if `GOOGLE_API_KEY` is not set.
`AUTOMERGENT_PROVIDER`	Core	Override the configured provider (`google-aistudio`, `google-vertex`, `custom`).
`AUTOMERGENT_MODEL`	Core	Override the configured model (e.g., `gemini-3.6-flash`).
`AUTOMERGENT_MODE`	Core	Override the operating mode (e.g., `edit`, `code`, `chat`, `agent`).

Note: Provider-specific API keys are always read from environment variables first, even when configured in the YAML file via "${ENV_VAR}" syntax. This ensures secrets are never persisted in config files.

CLI Flags

CLI flags provide one-shot overrides. They have the highest precedence and are never persisted to config files.

Flag	Argument	Description
`--provider`	`<name>`	AI provider to use.
`--model`	`<name>`	Model identifier.
`--mode`	`<mode>`	Operating mode (`edit`, `code`, `chat`, `agent`).
`--prompt`	`<text>`	Send a single prompt in non-interactive mode.
`--output`	`<format>`	Output format (`text`, `json`).
`--no-tui`	flag	Run in CLI-only mode without the terminal UI.
`--stdin`	flag	Read prompt from stdin.
`--theme`	`<name>`	UI theme to use.
`--keybindings`	`<preset>`	Keybinding preset (`default`, `vim`, `emacs`).
`--config`	`<path>`	Path to config file (overrides default locations).
`--no-color`	flag	Disable ANSI color output.
`--session`	`<id>`	Session ID to resume.
`--resume`	flag	Resume the most recent session.
`--new`	flag	Force a new session (don't resume).
`--session-dir`	`<path>`	Custom session storage directory.
`--context-files`	`<paths>`	Extra files to inject into context (comma-separated).
`--images`	`<paths>`	Image files to include as context.
`--files`	`<paths>`	Files to read into context at startup.
`--no-context`	flag	Disable automatic context gathering.
`--no-sandbox`	flag	Disable the sandbox (sets `sandbox: off`).
`--sandbox`	`<mode>`	Sandbox mode (`off`, `auto`, `strict`).
`--quiet`	flag	Suppress non-essential output.
`--verbose`	flag	Enable verbose output.
`--no-animation`	flag	Disable TUI animations.
`--layout`	`<mode>`	UI layout (`default`, `minimal`, `full`, `split`).
`--base-url`	`<url>`	Custom API endpoint URL for the active provider.
`--api-key`	`<key>`	API key for the active provider.

Example Configurations

Minimal Config

The simplest config — just set your provider and API key:

config.yaml
# Minimal configuration — everything else uses defaults
provider: "google-aistudio"
model: "gemini-3.6-flash"
mode: "manual"

providers:
  google-aistudio:
    apiKey: "${GEMINI_API_KEY}"

Multi-Provider Setup

Configure multiple providers with automatic fallback:

config.yaml
# Multi-provider setup with fallback chain
provider: "google-aistudio"
model: "gemini-2.5-pro"
mode: "manual"
theme: "catppuccin"

providers:
  google-aistudio:
    apiKey: "${GEMINI_API_KEY}"
    temperature: 0.2
    maxRetries: 3

  google-vertex:
    project: "my-gcp-project"
    location: "us-central1"
    temperature: 0.2

  custom:
    apiKey: "${PROVIDER_API_KEY}"
    baseUrl: "https://api.openai.com/v1"
    defaultModel: "gpt-4o"
    temperature: 0.3

fastModel: "gemini-3.6-flash"
effort: "medium"

providerFallback:
  - provider: "google-aistudio"
    model: "gemini-3.6-flash"
  - provider: "google-vertex"
    model: "gemini-3.6-flash"

Enterprise Config with Security

Harden Automergent for production/enterprise environments:

config.yaml
# Enterprise configuration with strict security
provider: "google-vertex"
model: "gemini-2.5-pro"
mode: "manual"
theme: "modern"

providers:
  google-vertex:
    project: "enterprise-ai-project"
    location: "us-central1"
    backend: "vertex"
    maxRetries: 5
    timeoutSeconds: 120
    temperature: 0.1
    headers:
      X-Request-Source: "automergent-enterprise"

# Security hardening
security:
  sandbox: "strict"
  requireGitForAutoModes: true
  blockedWritePaths:
    - ".git/**"
    - "~/.ssh/**"
    - "~/.gnupg/**"
    - "~/.aws/**"
    - "~/.kube/**"
    - "/etc/**"
    - "/var/**"
    - "/usr/**"
  allowedWritePaths:
    - "src/**"
    - "tests/**"
    - "docs/**"
  stripEnvVarPatterns:
    - "*_SECRET"
    - "*_PASSWORD"
    - "*_TOKEN"
    - "*_KEY"
    - "AWS_*"
    - "GCP_*"

# Tool configuration
tools:
  shell:
    enabled: true
    confirmationRequired: "always"
    timeout: "60s"
  filesystem:
    enabled: true
    confirmationRequired: "write"

# Review all edits before applying
editReview: true

# Diagnostics
diagnostics:
  enabled: true
  showInRead: true
  blockOnError: true
  blockOnWarning: true

# Privacy
zeroDataRetention: true
telemetry: false

# Git co-authoring
git:
  coAuthor: "always"
  coAuthorTrailer: "Co-authored-by: AI Assistant <ai@company.com>"

# Hooks for CI integration
hooks:
  postToolUse:
    - name: "lint"
      command: "make lint"
      timeout: "60s"

Performance Tuning

Optimize for large codebases and fast iteration:

config.yaml
# Performance-tuned configuration
provider: "google-aistudio"
model: "gemini-2.5-pro"
mode: "code"

# Maximize context for large codebases
maxContextTokens: 1000000
maxToolOutputChars: 64000
warnAtContextFraction: 0.85
autoCompressAt: 0.95
compressionKeepRecent: 20

# File handling for large repos
maxAutoReadFileSize: 1048576  # 1 MB
maxTreeFiles: 5000
maxTreeDepth: 15
excludePatterns:
  - ".git"
  - "node_modules"
  - "dist"
  - "build"
  - ".next"
  - "target"
  - "vendor"
  - "*.min.js"
  - "*.min.css"
  - "*.bundle.js"

# Provider tuning
providers:
  google-aistudio:
    apiKey: "${GEMINI_API_KEY}"
    temperature: 0.3
    maxTokens: 8192
    maxRetries: 3
    timeoutSeconds: 60

fastModel: "gemini-3.6-flash"
effort: "low"
thinkingBudget: 4096

# Disable features you don't need
notifications:
  desktop: false
  bell: false
  contextWarning: true

# Keep sessions lightweight
autoSave: true
checkpointInterval: 3
maxSessions: 50
maxSessionAge: "14d"

# Prompt caching for repeated operations
cache:
  prompt:
    enabled: true

Full Example Config

A comprehensive config showing all available options with their defaults:

config.yaml — Complete Reference
# ─── Core ───────────────────────────────────────────────────────
provider: "google-aistudio"
model: "gemini-3.6-flash"
mode: "manual"
theme: "modern"
keybindings: "default"
layout: "default"

# ─── Session ────────────────────────────────────────────────────
autoSave: true
checkpointInterval: 5
sessionDir: "~/.automergent/sessions"
maxSessions: 100
maxSessionAge: ""  # no age limit

# ─── Context ────────────────────────────────────────────────────
maxContextTokens: 128000
maxToolOutputChars: 32000
warnAtContextFraction: 0.8
autoCompressAt: 0.9
compressionKeepRecent: 10

# ─── Model ──────────────────────────────────────────────────────
fastModel: ""  # empty = use main model
thinkingBudget: 0  # 0 = provider default
effort: ""  # minimal|low|medium|high|max

# ─── File Handling ──────────────────────────────────────────────
maxAutoReadFileSize: 524288  # 512 KB
maxTreeFiles: 1000
maxTreeDepth: 10
excludePatterns:
  - ".git"
  - "node_modules"
  - ".venv"
  - "__pycache__"
  - "*.pyc"
  - "*.o"
  - "*.a"
  - "vendor"
contextFiles: []
skillsDir: "~/.automergent/skills"

# ─── Display ────────────────────────────────────────────────────
noAnimation: false
noColor: false
output: "text"
quiet: false
verbose: false

# ─── Edit ───────────────────────────────────────────────────────
editReview: false

# ─── Security ───────────────────────────────────────────────────
security:
  sandbox: "auto"
  requireGitForAutoModes: true
  rootRiskAcknowledged: false
  blockedWritePaths:
    - ".git/**"
    - "~/.ssh/**"
    - "~/.gnupg/**"
    - "~/.aws/**"
    - "~/.kube/**"
  allowedWritePaths: []
  stripEnvVarPatterns:
    - "*_SECRET"
    - "*_PASSWORD"
    - "*_TOKEN"
    - "*_KEY"

# ─── Tools ──────────────────────────────────────────────────────
tools:
  shell:
    enabled: true
    confirmationRequired: "always"
    timeout: "30s"
    maxOutputBytes: 1048576
  filesystem:
    enabled: true
    confirmationRequired: "write"
    maxOutputBytes: 524288
  git:
    enabled: true
    confirmationRequired: "never"
  web:
    enabled: true
    confirmationRequired: "never"
    timeout: "15s"

# ─── LSP ────────────────────────────────────────────────────────
lsp:
  enabled: true
  servers: {}
  startupTimeout: "10s"
  requestTimeout: "5s"

# ─── MCP ────────────────────────────────────────────────────────
mcp:
  servers: {}

# ─── Logging ────────────────────────────────────────────────────
log:
  level: "warn"
  file: "~/.automergent/automergent.log"
  maxSize: "50MB"
  maxBackups: 3

# ─── Providers ──────────────────────────────────────────────────
providers:
  google-aistudio:
    apiKey: "${GEMINI_API_KEY}"

# ─── Git ────────────────────────────────────────────────────────
git:
  coAuthor: "ask"
  coAuthorTrailer: "Co-authored-by: Automergent <automergent-bot@users.noreply.github.com>"

# ─── Diagnostics ────────────────────────────────────────────────
diagnostics:
  enabled: true
  showInRead: true
  blockOnError: true
  blockOnWarning: false
  maxFileSizeBytes: 1048576
  cacheDurationSec: 30

# ─── Cache ──────────────────────────────────────────────────────
cache:
  prompt:
    enabled: true

# ─── Privacy ────────────────────────────────────────────────────
zeroDataRetention: false
telemetry: true
noUpdateCheck: false

# ─── Debug ──────────────────────────────────────────────────────
debug:
  enabled: true
  directory: "~/.automergent/debug"
  saveRequests: true
  saveResponses: true
  maxFileSize: 10

# ─── Provider Fallback ──────────────────────────────────────────
providerFallback: []

# ─── Notifications ──────────────────────────────────────────────
notifications:
  desktop: false
  bell: true
  contextWarning: true

# ─── Hooks ──────────────────────────────────────────────────────
hooks:
  preToolUse: []
  postToolUse: []

# ─── Timeouts ───────────────────────────────────────────────────
confirmationTimeout: "10m"

Automergent Configuration Reference — GitHub — Docs