claude-code-agent/agent.conf.example at main · gregertw/claude-code-agent · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
# =============================================================================
# Agent Server Configuration
# =============================================================================
# Copy this file to agent.conf and fill in your values before running deploy.sh.
#
# This file is read during initial deployment only. deploy.sh copies it to the
# server as ~/.agent-server.conf, where it provides:
#   - OWNER_NAME    — used in the orchestrator's Claude prompt
#   - OUTPUT_FOLDER  — output path (output/tasks/, output/logs/, etc.)
#   - SETUP_*        — which MCP servers to verify in agent-setup.sh
#
# Schedule mode is managed separately via ~/.agent-schedule on the server
# (use agent-manager.sh --scheduled/--always-on to change it).
#
# After deployment, changing this file locally has NO effect on the server.
# To change server config, edit ~/.agent-server.conf on the server directly.
# =============================================================================

# --- Identity ----------------------------------------------------------------
# Your name (used in prompts and logging)
OWNER_NAME="Your Name"

# --- Paths -------------------------------------------------------------------
# Subfolder inside the brain where final outputs and documents go.
# This is also where the INBOX folder lives.
OUTPUT_FOLDER="output"

# --- Web Terminal ------------------------------------------------------------
# Install ttyd for browser-based terminal access? (true/false)
# Provides a web terminal at https://<ip> with self-signed TLS certificate.
# Useful as an alternative or supplement to SSH.
INSTALL_TTYD=false

# Username and password for ttyd web terminal (required if INSTALL_TTYD=true)
# IMPORTANT: Change these from the defaults!
TTYD_USER="agent"
TTYD_PASSWORD="changeme"

# --- Schedule ----------------------------------------------------------------
# "always-on" — instance runs 24/7, use cron or manual runs
# "scheduled" — instance wakes on a schedule, runs tasks, stops itself
SCHEDULE_MODE="always-on"

# Minutes between wakeups (only used in scheduled mode)
SCHEDULE_INTERVAL=60

# Active hours for cron (cron hour range syntax, e.g. "6-22" or "*" for all)
SCHEDULE_HOURS="6-22"

# Weekend hours — comma-separated hours to run on Sat/Sun (e.g. "8,17")
# When set, weekdays use SCHEDULE_HOURS + SCHEDULE_INTERVAL, weekends fire
# only at the listed hours. Leave empty to use the same schedule every day.
#SCHEDULE_WEEKEND_HOURS=""

# --- AWS ---------------------------------------------------------------------
# AWS region for the EC2 instance and EventBridge Scheduler
# Pick the one from AWS regions that is closest to you
AWS_REGION="eu-north-1"

# --- Agent Code Upgrade ------------------------------------------------------
# Automatically pull latest agent code from GitHub on boot and daily?
# Runs: agent upgrade (updates scripts and templates, not brain/ files)
AUTO_UPGRADE=true

# --- Installation Options ----------------------------------------------------
# Linux username on the EC2 instance
UBUNTU_USER="ubuntu"

# --- MCP Connections ---------------------------------------------------------
# ActingWeb Personal AI Memory (required — provides cross-session memory)
# Account is automatically created on first /mcp authentication
# Authentication is handled via OAuth during agent-setup.sh
SETUP_ACTINGWEB=true

# Gmail MCP (optional — enables email triage)
# Uses Anthropic's hosted connector at https://gmail.mcp.claude.com/mcp
# Authentication is via interactive Google sign-in (no credentials needed)
SETUP_GMAIL=false

# Google Calendar MCP (optional — enables calendar preview)
# Uses Anthropic's hosted connector at https://gcal.mcp.claude.com/mcp
# Authentication is via interactive Google sign-in (no credentials needed)
SETUP_GOOGLE_CALENDAR=false

# --- Claude CLI ---------------------------------------------------------------
# These settings control the flags passed to 'claude -p' in the orchestrator
# and run-agent.sh scripts.

# Model to use (e.g. "sonnet", "opus", "haiku", or a full model ID).
# Leave empty to use Claude Code's default.
#CLAUDE_MODEL=""

# Maximum conversation turns per run (prevents runaway loops)
CLAUDE_MAX_TURNS=50

# Output format: "text", "json", or "stream-json"
CLAUDE_OUTPUT_FORMAT="text"

# Permission mode: "default", "acceptEdits", "bypassPermissions", "plan", "auto"
# For autonomous server runs, "bypassPermissions" is typical.
#CLAUDE_PERMISSION_MODE=""

# Maximum dollar budget per run (leave empty for unlimited)
#CLAUDE_MAX_BUDGET_USD=""

# Enable verbose output (true/false)
CLAUDE_VERBOSE=false

# Additional CLI flags passed verbatim to claude (advanced, use with care)
#CLAUDE_EXTRA_FLAGS=""