Augureaugure

CLI Reference

Augure CLI reference — all commands for starting, stopping, configuring, and managing your AI agent from the terminal

Augure ships a CLI (augure) for managing the agent lifecycle, skills, memory, jobs, and more.

Install

npm install -g augure

Commands

augure init

Scaffolds a new Augure project in the current directory:

augure init

Creates two files:

  • augure.json5 -- default configuration (see Configuration)
  • .env -- template for secret environment variables

augure start

Starts the agent:

augure start                           # foreground, uses ./augure.json5
augure start -c /path/to/config        # custom config path
augure start --daemon                  # run as background daemon
augure start --mcp                     # enable MCP server
augure start --daemon --mcp            # daemon + MCP
FlagAliasDefaultDescription
--config-c./augure.json5Path to config file
--env-e.env next to configPath to .env file
--debugfalseEnable debug logging
--daemonfalseRun as background daemon
--mcpfalseEnable MCP HTTP server

In daemon mode, the agent runs as a detached process. Logs are written to augure.log next to the config file. A PID file (.augure.pid) is written for lifecycle management.

augure stop

Stop a running agent:

augure stop                            # stop agent using default config location
augure stop -c /path/to/config         # stop agent for specific config

Reads the .augure.pid file, sends SIGTERM, waits up to 5 seconds for graceful shutdown, then force-kills if needed.

augure status

Show an overview of the agent's state:

augure status
▲ augure status
  Agent:    running (PID 12345)
  Identity: Augure
  LLM:      anthropic/claude-sonnet-4-5 (openrouter)
  Memory:   ./memory (14 files)
  Jobs:     5 active, 1 disabled
  Channels: telegram ✓  whatsapp ✗  web ✗
  Docker:   connected
  Skills:   12 found

augure doctor

Check configuration and connectivity:

augure doctor
augure doctor -c /path/to/config

Runs 4 checks:

  1. Config validation -- loads and validates augure.json5 against the Zod schema
  2. Docker connectivity -- pings the Docker daemon
  3. LLM connectivity -- verifies API key with a lightweight /models request
  4. Telegram connectivity -- validates bot token with getMe()
▲ augure doctor

  ✓ Config valid
  ✓ Docker connected
  ✓ LLM reachable (openrouter)
  ✓ Telegram bot connected (@my_bot)

augure memory

Manage agent memory files. All subcommands accept --config, -c.

augure memory list [directory]

List memory files with sizes:

augure memory list
augure memory list tasks/

augure memory show <path>

Print the content of a memory file:

augure memory show notes.md

augure memory edit <path>

Open a memory file in $EDITOR (fallback: vi):

augure memory edit notes.md

augure jobs

Manage scheduled jobs. All subcommands accept --config, -c.

augure jobs list

List all jobs (from config + persisted):

augure jobs list
ID                    TYPE      SCHEDULE              ENABLED  CHANNEL     LAST RUN
morning-check         cron      0 8 * * *             yes      telegram    2026-02-25T08:00:00Z
remind-deadline       one-shot  2026-03-15T09:00      yes      telegram    (pending)

augure jobs add

Add a new persisted job:

augure jobs add --id daily-news --cron "0 9 * * *" --prompt "Summarize today's tech news"
augure jobs add --id reminder --runAt "2026-03-15T09:00:00Z" --prompt "Remind about deadline"

augure jobs remove <id>

Remove a persisted job:

augure jobs remove daily-news

augure jobs run <id>

Execute a job's prompt interactively (creates a minimal agent, runs the prompt, prints the response):

augure jobs run morning-check

augure channels

augure channels status

Show the status of each configured channel:

augure channels status

For Telegram, validates the bot token online. Other channels show enabled/disabled status.

augure tools

augure tools list

List all tools with their configuration status:

augure tools list
TOOL                  STATUS            RISK
memory_read           configured        -
memory_write          configured        -
sandbox_exec          configured        high
web_search            not configured    -
browser               not configured    -

augure tools test <name>

Test a tool with sample or custom input:

augure tools test datetime
augure tools test memory_read
augure tools test web_search --args '{"query":"test"}'

augure skills

Manage skills from the command line. All subcommands accept --config, -c.

augure skills list

List all skills with their status, version, trigger, and last update date.

augure skills show <id>

Display full details for a skill, including metadata and the markdown body.

augure skills run <id>

Trigger a skill execution manually in a Docker container.

augure skills pause <id>

Pause a skill (skipped by the cron scheduler).

augure skills resume <id>

Resume a previously paused skill.

augure skills delete <id>

Permanently delete a skill and all its files.

augure skills logs <id>

Show the last 10 execution runs for a skill.

augure mcp

Start the MCP server in stdio mode (for Claude Desktop, Cursor, etc.):

augure mcp -c ./augure.json5

See MCP Server for full documentation.

Skill ID Format

All skill IDs must be lowercase alphanumeric with hyphens (e.g. apartment-search, daily-digest). This is enforced by the CLI to prevent path traversal.

  • MCP Server -- expose Augure tools to Claude Desktop and other MCP clients
  • Skills System -- how skills are created, tested, and self-healed
  • Configuration -- the full augure.json5 reference
  • Scheduler -- cron triggers for skills

On this page