CLI Reference¶
Command reference for shipped lazyai-cli commands. Repository scripts such as bin/doctor, bin/inject, and bin/startup-self-heal are maintainer harness tools, not CLI commands; archived Fortnite/orchestrator/eval material is historical or migration context only.
The commands under ops-runtime-extra are still physically shipped today, but they are transitional extras outside the default setup-core product boundary rather than the headline LazyAI story. For the source-backed command category inventory and rationales, see Product Boundaries; the same category names are guarded by packages/cli/cmd/command_category.go.
Top-level command surface¶
These are the top-level commands registered by source under packages/cli/cmd/:
| Category | Commands |
|---|---|
setup-core |
add, build-plugin, compile, completion, config, create, doctor, eject, import, info, init, list, migrate, server, setup, sidecar, status, update, update-self, validate |
ops-runtime-extra |
auth, backup, cost, git, ledger, memory, message, metrics, notify, restore-runtime-db, secret, session |
dev-harness |
models |
retired/archived |
completions (hidden deprecated alias for completion) |
Removed command surfaces such as task, workflow, orchestration, mcp-setup, and obsolete eval are not active CLI commands.
Table of Contents¶
- Top-level command surface
- Init
- Add
- Compile
- Update
- Status
- Doctor
- Server
- server list
- server add
- server remove
- server doctor
- Config
- Create
- Import
- Migrate
- Eject
- Info
- List
- Build Plugin
- Update Self
- Setup
- Models
- Session
- Auth
- Cost
- Ledger
- Validate
- validate agents
- validate skills
- validate evals
- Migration Note
- Message
- Metrics
- Memory
- Completion
- Git
- Backup
- Restore Runtime DB
- Secret
- Notify
- Sidecar
- Environment Variables
- Exit Codes
Init¶
init¶
Initialize the AI development environment.
Flags:
| Flag | Description |
|---|---|
--scope |
Setup scope (project, global, workspace) |
--workspace-root |
Root directory for workspace-scoped installs (workspace scope) |
--tools |
Tools to configure (opencode, claude-code, copilot, pi, omp, kiro, antigravity) |
--preset |
Preset configuration (minimal, standard, full, custom) |
--enable-servers |
MCP servers to enable (for example: filesystem, ai-memory, ripgrep) |
--features |
Comma-separated features to enable |
--disable-features |
Comma-separated features to disable |
--name |
Project name |
--branch-pattern |
Git branch naming pattern |
--commit-pattern |
Git commit message pattern |
--existing-setup-policy |
How to handle existing setup (absorb, adapt, backup-only) |
--no-interactive |
Run without interactive prompts |
--drive-cli |
Delegate scaffolding to the tool's own CLI when available (Claude Code) |
--local-secrets |
Route Claude Code MCP/settings writes to gitignored .claude/settings.local.json instead of committed surfaces |
--org |
Organization name (populates [YOUR_ORG] in AGENTS.md) |
--team |
Team name (populates [YOUR_TEAM] in AGENTS.md) |
--force |
Overwrite existing files |
--dry-run |
Preview changes without writing |
--memory-path |
Project memory path (default: .specify/memory) |
--reversa |
Analyze existing code with Scout/Reversa to auto-populate project details |
--no-reversa |
Skip Scout/Reversa analysis and leave project details explicit/manual |
--express |
Run interactive wizard with Express mode |
--custom |
Run interactive wizard with Personalized mode |
OpenCode default behavior:
When OpenCode is selected, init installs the neutral canonical adapter path. Root opencode.json keeps the baseline config shape, .opencode/agents/guide.md is emitted as the front-door default agent, and Fortnite agents, .opencode/STARTUP.md, and loop-driver are not installed by default.
Examples:
# Interactive setup
lazyai-cli init
# Non-interactive with OpenCode
lazyai-cli init --tools opencode --preset standard --no-interactive
AGENTS.md placeholders (headless init):
init writes a canonical AGENTS.md that contains <!-- fill-in: ... --> markers for project-specific sections (organization, conventions, architecture, etc.). The CLI does not fill these markers itself because it cannot run your AI tool. After init finishes:
- Host tool path (recommended): Open the project in your AI tool (Claude Code, OpenCode, etc.) and run
/initor/populate. The AI tool indexes the codebase and replaces each marker with a value backed by code evidence. - Manual path: Edit
AGENTS.mdby hand and replace each<!-- fill-in: ... -->marker with a concrete value, or remove the marker if the section does not apply. - Validate after populate:
lazyai-cli validate agents # confirm agent frontmatter is still valid
lazyai-cli doctor # full setup health check
Both --no-interactive and the interactive wizard leave markers unfilled on purpose; filling requires the AI tool to run.
Add¶
add¶
Add agents, skills, or tool configurations to an existing initialized setup, then rerun the scaffold pipeline and update the setup store.
Flags:
| Flag | Description |
|---|---|
--tools |
Tool IDs to add configuration for |
--agents |
Agent IDs to add |
--skills |
Skill IDs to add |
--no-interactive |
Run without prompts; at least one of --tools, --agents, or --skills is required |
Tool validation:
--tools is registry-backed. Current tool IDs are opencode, claude-code, copilot, pi, omp, kiro, and antigravity.
Examples:
lazyai-cli add --tools claude-code,pi --no-interactive
lazyai-cli add --agents reviewer,evidence-verifier --skills fast-feedback --no-interactive
Compile¶
compile¶
Compile .ai/mcp.json or .ai/mcp.jsonc into per-tool MCP/config files for the selected tools in the setup store.
Flags:
| Flag | Description |
|---|---|
--tool |
Compile only one tool (opencode, claude-code, copilot, pi, omp, kiro, or antigravity) |
--dry-run |
Preview the tools that would be compiled without writing |
--local-secrets |
Route Claude Code MCP writes to .claude/settings.local.json instead of committed config |
--validate-contracts |
Validate skill contracts before compile (default: true) |
--strict-contracts |
Turn contract warnings into compile failures |
Example:
Update¶
update¶
Refresh managed files to match the embedded library version while applying LazyAI conflict handling.
Flags:
| Flag | Description |
|---|---|
--force |
Overwrite local changes without prompting |
--no-interactive |
Run without interactive prompts |
--dry-run |
Show what would change without writing |
Example:
Status¶
status¶
Show current setup state, selected tools, file health, git conventions, and CLI version.
Flags:
| Flag | Description |
|---|---|
--dir |
Target directory (default: current directory) |
--json |
Output JSON |
Example:
Server¶
Manage MCP servers from the embedded catalog and the current setup store. The current catalog includes ai-memory, filesystem, ripgrep, codegraph, and obsidian.
server list¶
List cataloged MCP servers and show whether each is enabled.
Flags:
| Flag | Description |
|---|---|
--json |
Output JSON |
--dir |
Project directory (defaults to current directory) |
Example:
server add¶
Enable a cataloged MCP server in the setup store. Run lazyai-cli compile afterward to regenerate per-tool configs.
Arguments:
name(required): Catalog server name
Flags:
| Flag | Description |
|---|---|
--no-interactive |
Skip confirmation prompt |
--dir |
Project directory (defaults to current directory) |
Example:
Unknown server names fail with unknown MCP server: <name> and include the available catalog names in the error metadata.
server remove¶
Disable a cataloged MCP server in the setup store. Run lazyai-cli compile afterward to remove it from generated per-tool configs.
Arguments:
name(required): Catalog server name
Flags:
| Flag | Description |
|---|---|
--dir |
Project directory (defaults to current directory) |
--no-interactive |
Skip confirmation prompt |
Example:
server doctor¶
Validate enabled MCP server definitions in .ai/mcp.json and the generated per-tool MCP config files. Pass a server name to validate only one server.
L1 config check: Verifies the server entry exists in .ai/mcp.json, is enabled, and is present in each per-tool compiled MCP config file.
L3 stdio handshake (not performed): The Go binary does not bundle an MCP client library for the JSON-RPC handshake protocol. A TypeScript wrapper using @modelcontextprotocol/sdk could perform L3 checks; this is future work.
Arguments:
name(optional): Catalog server name
Flags:
| Flag | Description |
|---|---|
--json |
Output JSON |
--dir |
Project directory (defaults to current directory) |
Example:
Config¶
config init¶
Create .opencode/config.yaml with defaults for agent, model, database, ledger, and notifications.
config get¶
Read one config key.
config set¶
Set one config key. Critical keys print a warning before saving.
Supported keys: agent.default_agent, agent.default_model, database.path, ledger.path, notifications.enabled, and notifications.webhook.
config list¶
Print all current config values.
Create¶
create¶
Generate a new artifact using the artifact generator registry.
Usage:
Valid types: agent, skill, prompt, command, template, hook
Flags:
| Flag | Description |
|---|---|
--force |
Overwrite an existing artifact |
--no-interactive |
Require type and name from arguments; do not prompt |
--description |
Description used by generators that support it |
Example:
Invalid types fail with:
Import¶
import¶
Import configuration from another AI tool setup into LazyAI-managed form. If source is omitted, the current directory is scanned.
Usage:
Flags:
| Flag | Description |
|---|---|
--tool |
Source tool to import (opencode, claude-code, copilot, pi, antigravity, omp, or kiro); auto-detected if omitted |
--no-interactive |
Run without confirmation prompts |
--preview |
Show the import plan without writing |
--strategy |
Merge strategy: smart, preserve, replace, or append |
--verbose |
Show detailed output |
--skip-backup |
Do not create a backup before writing |
Example:
Migrate¶
migrate¶
Migrate an older LazyAI or ai-setup configuration into the current canonical format. If source is omitted, the current directory is scanned.
Usage:
Flags:
| Flag | Description |
|---|---|
--from |
Source version hint |
--no-interactive |
Run without confirmation prompts |
--preview |
Show the migration plan without writing |
--strategy |
Merge strategy: smart, preserve, replace, or append |
--verbose |
Show detailed output |
--skip-backup |
Do not create a backup before writing |
Example:
Eject¶
eject¶
Remove LazyAI management metadata so LazyAI stops managing the generated files. The generated files remain in place. The following files are removed if present: .ai/lazyai.json, .ai/lock.json, .ai/migration-report.md, .ai-setup.json, .ai-setup.db.
Flags:
| Flag | Description |
|---|---|
--no-interactive |
Skip confirmation prompt |
--dir |
Project directory (defaults to current directory) |
Example:
Info¶
info¶
Show metadata for one tracked artifact by basename or tracked path.
Arguments:
name(required): Artifact basename or tracked file path
Flags:
| Flag | Description |
|---|---|
--dir |
Project directory (defaults to current directory) |
--json |
Output JSON |
Example:
List¶
list¶
List installed tracked artifacts, optionally filtered by category.
Usage:
Flags:
| Flag | Description |
|---|---|
--type |
Filter by artifact type (agents, skills, templates, rules, and so on) |
--verbose |
Include file paths |
--json |
Output JSON |
--dir |
Project directory (defaults to current directory) |
Example:
Build Plugin¶
build-plugin¶
Generate a distributable plugin bundle from the embedded LazyAI library.
Supports four targets via --target: Claude Code (default), GitHub Copilot CLI,
OMP, and Pi.
Flags:
| Flag | Description |
|---|---|
--target |
Bundle target: claude (default), copilot-cli, omp, or pi |
--out |
Output directory (default: ./dist/plugin) |
--force |
Overwrite the output directory if it exists and is non-empty |
Example:
lazyai-cli build-plugin --target claude --out ./dist/lazyai-plugin --force
lazyai-cli build-plugin --target copilot-cli --out ./dist/copilot-plugin
Update Self¶
update-self¶
Download and replace the current lazyai-cli binary from GitHub Releases. Development builds print Development build — skipping self-update unless --version is provided.
Flags:
| Flag | Description |
|---|---|
--check |
Only check whether an update is available; exits non-zero when one is available |
--force |
Skip version comparison |
--dry-run |
Preview without downloading |
--verbose |
Show detailed release and asset information |
--version |
Download a specific release tag |
Example:
Setup¶
setup¶
Inspect setup inventory and setup planning output. This command reports setup state; it does not start daemons or runtime execution.
Flags:
| Flag | Description |
|---|---|
--scan |
Scan known tool targets and print inventory JSON |
--list |
List supported setup targets and reusable setup resources |
--dry-run |
Show the setup plan without writing |
--adopt |
Mark adoptable external configs as LazyAI managed; requires --scan |
--import |
Import external configs into LazyAI reference storage; requires --scan |
--tool |
Limit planning to a tool; repeatable and only valid with --list or --dry-run |
--all |
Select all supported setup targets for the requested scope |
--global |
Use global scope/home layout where supported |
Select exactly one of --scan, --list, or --dry-run.
Examples:
lazyai-cli setup --scan
lazyai-cli setup --list --tool claude-code
lazyai-cli setup --dry-run --global --all
Models¶
models sync¶
Maintainer-only dev-harness command that refreshes packages/cli/internal/models/catalog_gen.go from models.dev/api.json.
It filters the upstream catalog to the providers each supported CLI uses, checks curated tier references, prints a diff, and writes only after confirmation unless --yes is passed.
Flags:
| Flag | Description |
|---|---|
--yes |
Skip the confirmation prompt and write if there are no missing curated references |
--output |
Write the regenerated catalog to a custom path |
--dry-run |
Print the diff without writing |
Example:
Session¶
session start [goal]¶
Start a new AI agent session.
Arguments:
- goal (required): Brief description of the session goal
Example:
Output:
✅ Session started: ses_1234567890
Goal: Implement authentication feature
Started: 2026-05-24T00:08:16Z
session list¶
List all sessions.
Example:
Output:
Sessions:
---------
🟢 ses_1234567890 | Implement auth | 2026-05-24T00:08:16Z
🔴 ses_1234567889 | Fix bug #42 | 2026-05-23T23:45:00Z
session show [session-id]¶
Show session details.
Arguments:
- session-id (required): Session ID (e.g., ses_1234567890)
Example:
session end [session-id]¶
End a session.
Arguments:
- session-id (required): Session ID to end
Example:
Doctor¶
doctor¶
Validate setup integrity and run six environment health checks.
Integrity checks:
- Managed files exist and match tracked hashes
- Stray
AGENTS.mdfiles underspecs/ - Metadata gaps in spec frontmatter
Environment health checks:
sqlite3dependency — fail if not foundgitdependency — fail if not foundjqdependency — warn if not foundbashdependency — fail if not foundollamaprovider — warn if not running onlocalhost:11434- Disk space — warn above 80% used, fail above 90% used
Flags:
| Flag | Description |
|---|---|
--dir |
Project directory to check (defaults to current directory) |
--fix |
Print fix instructions for detected issues |
--verbose |
Show detailed output for all files and metadata issues |
--json |
Output JSON |
Example:
Output:
🩺 Integrity Check
Status ✅ All files healthy
Health ████████████████████ 100%
Total files 42
Healthy 42
Missing 0
Modified 0
Stray AGENTS.md 0
Metadata gaps 0
Health warnings 2
🩺 Environment Health Checks
✅ Dependency: sqlite3 | 3.39.5
✅ Dependency: git | git version 2.39.0
⚠️ Dependency: jq | jq not found (optional but recommended)
✅ Dependency: bash | GNU bash, version 5.2.0
⚠️ Provider: ollama | Ollama not running on localhost:11434
✅ Disk space | 45% used
Auth¶
auth list¶
List configured AI providers from the setup store and show common authentication environment variables with masked values.
Example:
Checks OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY, and GITHUB_TOKEN.
Cost¶
cost show¶
Show cost breakdown by session.
Flags:
| Flag | Description |
|---|---|
--period, -p |
Time period: day, week, month, or all |
cost agent¶
Show cost and dispatch counts grouped by agent.
cost budget¶
Show current spend against an optional budget limit.
Flags:
| Flag | Description |
|---|---|
--limit, -l |
Budget limit in USD |
Ledger¶
ledger init¶
Initialize the immutable ledger.
Example:
ledger append [event-type] [data]¶
Append an event to the ledger.
Arguments:
- event-type (required): Type of event (e.g., dispatch, session_start)
- data (required): Event data as key=value pairs
Example:
ledger verify¶
Verify ledger integrity.
Example:
Output:
ledger show [count]¶
Show recent ledger entries.
Arguments:
- count (optional): Number of entries to show (default: 10)
Example:
Validate¶
The validate command runs structural validators over the canonical .ai/
asset tree. With no subcommand it prints help; pass --all to run every
validator in one pass, or use the agents, skills, and evals subcommands
for filtered per-surface checks.
Flags:
| Flag | Description |
|---|---|
--all |
Run all validators (skills, agents, hooks, MCP, secrets, path safety) over the canonical .ai/ tree and return an error on any error-severity issue |
--profile |
Safety profile: team (inline secrets are errors) or personal (inline secrets are warnings). Defaults to the manifest profile, then personal |
validate agents¶
Validate .ai/agents/*.md files using the consolidated validation engine.
The engine runs all validators over the canonical .ai/ tree and then filters
to agent-rule issues. Counts agent files as passed, surfaces error and warning
findings, and exits with an error only when error-severity checks fail.
Checks:
- YAML frontmatter is present (error)
namefield is present and valid (error)descriptionfield is present (warning)- Semantic body guidance for role, workflow, trigger, non-trigger, human gate, handoff, evidence, and output (warning)
Example:
validate skills¶
Validate .ai/skills/ using the consolidated validation engine over the
canonical .ai/ tree. Filters to skill-rule issues, counts pass/fail across
flat *.md files and <name>/SKILL.md directories, and returns an error on
any error-severity failure.
Checks:
- YAML frontmatter is present (error)
namefield is present and valid (error)descriptionfield is present and non-empty (error)- Semantic body guidance for trigger, non-trigger, evidence, and output (warning)
Example:
validate evals¶
Validate .ai/evals/cases/, .ai/evals/holdouts/, and .ai/evals/rubrics/.
Case and holdout files (.yaml/.yml) are schema-checked by the evals
package; rubric directories are validated for structure. Counts pass/fail and
returns an error on any failure.
Example:
Migration Note¶
The task, workflow, orchestration, mcp-setup, and obsolete eval command surfaces were removed during the runtime refactor.
Use:
docs/migration/fortnite-orchestrator-removal.mdfor user-facing migration and rollback guidancelazyai-cli create,lazyai-cli list, and adapter-managed files for current scaffolding flows
Do not use repository harness scripts or archived assets as replacements for removed CLI commands unless you are maintaining this repository itself.
Message¶
message send [to-agent] [subject] [body]¶
Send a message to an agent.
Arguments:
- to-agent (required): Target agent name
- subject (required): Message subject
- body (required): Message body
Flags:
- --priority, -p: Message priority (low, normal, high, critical)
Example:
message recv [agent]¶
Receive messages for an agent.
Arguments:
- agent (required): Agent name to receive messages for
Behavior: - Marks all unread messages for the agent as read - Shows the 10 most recent messages
Example:
message broadcast [subject] [body]¶
Broadcast a message to all agents.
Arguments:
- subject (required): Broadcast subject
- body (required): Broadcast body
Flags:
- --priority, -p: Message priority (low, normal, high, critical)
Example:
Metrics¶
metrics list¶
List recent quality metrics.
Flags:
- --limit, -n: Number of metrics to show (default: 10)
Example:
metrics export¶
Export metrics to Prometheus format.
Flags:
- --output, -o: Output file path (default: metrics.prom)
Safety note: Writes a file to the current directory. Existing files are overwritten.
Example:
metrics dashboard¶
Generate HTML dashboard.
Flags:
- --output, -o: Output file path (default: dashboard.html)
Safety note: Writes a file to the current directory. Existing files are overwritten.
Example:
Memory¶
memory save [content]¶
Save a memory.
Arguments:
- content (required): Memory content
Flags:
- --type, -t: Memory type (lesson, context, decision, idea)
- --tags: Tags for categorization
Example:
memory list¶
List all memories.
Example:
memory search [query]¶
Search memories.
Arguments:
- query (required): Search query
Example:
Completion¶
completion [shell]¶
Generate shell completion scripts.
Arguments:
- shell (required): One of bash, zsh, fish, powershell
Example:
# Bash
source <(lazyai-cli completion bash)
# Zsh
source <(lazyai-cli completion zsh)
# Fish
lazyai-cli completion fish | source
Git¶
git sync¶
Auto-commit all changes with descriptive messages.
Safety note: This command automatically stages (git add -A) and commits ALL changes in the current repository. It shows the list of files and asks for confirmation before proceeding. Use --force to skip the confirmation prompt.
Flags:
- --message, -m: Custom commit message
- --force: Skip confirmation prompt
Example:
git log¶
Show recent commits, highlighting LazyAI-attributed commits.
Flags:
- --limit, -n: Number of commits to show (default: 10)
Example:
git status¶
Show git status with LazyAI context, including the most recent active session when one is recorded.
Example:
Backup¶
backup create¶
Create a backup of LazyAI data.
Flags:
- --output, -o: Output file path (default: lazyai-backup-YYYYMMDD_HHMMSS.tar.gz)
Example:
backup restore [backup-file]¶
Restore LazyAI data from a backup tarball.
Arguments:
- backup-file (required): Path to the backup file
Safety note: This overwrites current LazyAI data with the backup contents. It asks for confirmation before proceeding. Use --force to skip the confirmation prompt.
Flags:
- --force: Skip confirmation prompt
Example:
Restore Runtime DB¶
restore-runtime-db¶
Restore .specify/session.db from a backup file. The current runtime database is renamed to .specify/session.db.pre-restore before the backup is copied into place.
Arguments:
backup-path(required): Path to the SQLite backup file
Flags:
| Flag | Description |
|---|---|
--force |
Skip confirmation prompt |
Safety note: This overwrites the current runtime database. Data written since the backup was created will be lost.
Example:
Secret¶
secret set [name] [value]¶
Store a secret.
Arguments:
- name (required): Secret name
- value (required): Secret value
Safety note: Secrets are stored using the OS keychain when available, or a fallback file in ~/.lazyai/secrets/ (not encrypted at rest). Do not use for production credentials.
Example:
secret get [name]¶
Retrieve a secret.
Arguments:
- name (required): Secret name
Safety note: The secret value is printed to stdout. Be careful when running this in shared environments or logging output.
Example:
secret list¶
List stored secrets.
Example:
secret remove [name]¶
Remove a secret.
Arguments:
- name (required): Secret name
Flags:
- --force: Skip confirmation prompt
Example:
Notify¶
notify config¶
Configure notification settings.
Flags:
- --webhook: Webhook URL for notifications
- --enabled: Enable or disable notifications
Safety note: The webhook URL is stored in the local config file. Ensure the URL is trusted and uses HTTPS.
Example:
notify send [message]¶
Send a desktop notification.
Arguments:
- message (required): Notification message
Flags:
- --title, -t: Notification title (default: "LazyAI")
Example:
notify test¶
Send a test notification to verify configuration.
Example:
Sidecar¶
Manage optional sidecar directories for docs, specs, and plans. Sidecar configuration is discovered positionally — LazyAI walks up from the current directory to find .lazyai/sidecar.yaml layers and merges them field-by-field. See Scopes → Sidecar and scope behavior for the full discovery/merge model. Remaining surface: init, status, doctor — there is no attach/detach and no separate workspace command; a "workspace" layer is just an ancestor directory's .lazyai/sidecar.yaml, created the same way as a project layer.
sidecar init¶
Initialize a sidecar config at the given scope. Writes <target>/.lazyai/sidecar.yaml.
Flags:
- --scope: Scope label (workspace, project, global). Default: project. workspace and project write to the identical path (cwd/.lazyai/sidecar.yaml) — the value is purely a label for the confirmation message and ledger entry, not a different write target; see Scopes doc. --scope global writes to ~/.lazyai/sidecar.yaml.
- --path: Sidecar root path (required). This is the sidecar's content root (where docs/specs/plans actually live) — independent of where sidecar.yaml itself is written.
- --specs-dir: Override specs directory name. Default: specs.
- --docs-dir: Override docs directory name. Default: docs.
- --plans-dir: Override plans directory name. Default: plans.
Examples:
# Project-scope sidecar (default when --scope is omitted)
lazyai-cli sidecar init --path ../shared-docs
# Workspace-scope sidecar — run from the shared ancestor directory;
# every project underneath it picks this up automatically
lazyai-cli sidecar init --scope workspace --path /Users/me/kb/my-workspace
# Global sidecar
lazyai-cli sidecar init --scope global --path ~/kb
sidecar status¶
Show every discovered .lazyai/sidecar.yaml layer (global, workspace, project) plus the resolved docs/specs/plans paths, with per-field provenance showing which layer supplied each value.
Flags: none
Example:
lazyai-cli sidecar status
# → Sidecar Status (cwd: /path/to/project)
# →
# → Layers discovered:
# → global /Users/me/.lazyai/sidecar.yaml (found)
# → workspace /path/to/.lazyai/sidecar.yaml (not found)
# → project /path/to/project/.lazyai/sidecar.yaml (found)
# →
# → Resolved paths:
# → docs_dir /path/to/project/docs (from: project)
# → specs_dir /path/to/project/specs (from: global)
# → plans_dir /path/to/project/plans (from: default)
If no layer is found anywhere, sidecar status additionally prints: (No .lazyai/ configuration found — using built-in defaults. Run 'sidecar init' to configure.)
sidecar doctor¶
Validate every discovered sidecar layer's path exists and is writable, and flag a legacy ~/.lazyai/workspaces.yaml if still present. Reports issues with exit codes.
Flags: none
Exit codes:
- 0: All paths valid, or warnings only (e.g., missing path that will be created on first write). WARN lines are printed but the command succeeds.
- 1: Errors found (e.g., non-writable directory, file where directory expected).
Example:
Each reported issue is prefixed with the layer it came from — [global], [workspace], or [project] — or [general] for issues not tied to a single layer (e.g. the legacy-registry warning):
lazyai-cli sidecar doctor
# → [project] ERROR: sidecar path does not exist: /path/to/project/kb
# → [general] WARN: found legacy ~/.lazyai/workspaces.yaml (workspace registry removed in #579) — re-create any workspace/project sidecars you still need with 'sidecar init' from the intended directory
Environment Variables¶
| Variable | Description | Default |
|---|---|---|
LAZYAI_AGENT |
Current agent name | implementer |
LAZYAI_DB_PATH |
Database file path | .specify/lazyai.db |
LAZYAI_LEDGER_PATH |
Ledger file path | .specify/ledger.jsonl |
Exit Codes¶
| Code | Meaning |
|---|---|
| 0 | Success |
| 1 | General error |
| 2 | Validation error |
| 3 | Database error |
| 4 | Ledger error |