-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy path.env.example
More file actions
159 lines (123 loc) · 6.25 KB
/
Copy path.env.example
File metadata and controls
159 lines (123 loc) · 6.25 KB
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
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
# ============================================================================
# My Agent — environment configuration
# ============================================================================
# Copy this file to `.env` and fill in the values you need.
# cp .env.example .env
#
# Precedence: CLI flags > environment variables > defaults.
# Every variable below is OPTIONAL unless marked [required].
# ============================================================================
# ----------------------------------------------------------------------------
# Model connection (two API styles)
# ----------------------------------------------------------------------------
# API style — determines which client protocol to use:
# openai — OpenAI-compatible chat completions (OpenAI, Ollama, DeepSeek, OpenRouter, gateways)
# anthropic — Anthropic Messages API
# Aliases: STYLE
#MODEL_STYLE=openai
# Model id sent to the provider API. [required — no default]
# Examples: gpt-4.1, qwen2.5-coder:7b, anthropic/claude-3.5-sonnet, deepseek-chat
#MODEL=qwen2.5-coder:7b
# API base URL — used as-is (trailing slashes are trimmed). No automatic /v1 suffix.
# OpenAI-compatible examples:
# https://api.deepseek.com (DeepSeek — no /v1)
# https://api.openai.com/v1
# http://localhost:11434/v1 (Ollama)
# Defaults per style when unset:
# openai → https://api.openai.com/v1
# anthropic → https://api.anthropic.com
# Aliases: MODEL_BASE_URL
#BASE_URL=http://localhost:11434/v1
# API key. Required for anthropic style; optional for local OpenAI-compatible endpoints.
#API_KEY=
# ----------------------------------------------------------------------------
# Agent runtime
# ----------------------------------------------------------------------------
# Maximum number of agent loop iterations per turn. Default: 50.
#MAX_ITERATIONS=50
# ----------------------------------------------------------------------------
# Sandbox (CLI / Node host)
# ----------------------------------------------------------------------------
# Controls how shell commands and filesystem access are sandboxed.
# local — OS-level sandbox (dropped privileges / restricted env) [default]
# native — no sandbox, runs directly in the host environment
#SANDBOX_ENV=local
# Suppress sandbox downgrade warning banners (set to 1 to silence).
#MY_AGENT_SANDBOX_QUIET=0
# ----------------------------------------------------------------------------
# Remote CoreEnv (workspace) server
# ----------------------------------------------------------------------------
# CLI: forwards fs/shell/env operations to a remote CoreEnv HTTP server instead
# of the local Node environment (`--remote-env` / `-R` or REMOTE_ENV).
# Server (pnpm start:server): registers the remote env as its own workspace, so
# a remote-session host can chain an even further CoreEnv.
# Example: http://localhost:3100
#REMOTE_ENV=http://localhost:3100
# Remote model provider (combines freely with REMOTE_ENV).
# CLI adapters call this server's /api/provider/*; the *server* injects API_KEY
# and must be able to reach its own BASE_URL (VPN/intranet routes on the server host).
#REMOTE_PROVIDER=http://localhost:3200
# Remote Agent Session — exclusive on the CLI: do NOT combine with REMOTE_ENV /
# REMOTE_PROVIDER on the same client (the agent loop runs server-side). Use
# --model <id> to push local LLM settings to the server; the server itself may
# register its own REMOTE_ENV / REMOTE_PROVIDER.
# CLI/extension/playground bind remote sessions through this server's /api/agent/*
# via @my-agent/server/agent-session (RemoteSessionClient).
#REMOTE_SESSION=http://localhost:3200
# ----------------------------------------------------------------------------
# CoreEnv HTTP server (pnpm start:server)
# ----------------------------------------------------------------------------
# Port the CoreEnv server listens on. Default: 3100.
#SERVER_PORT=3100
# Root directory the server is allowed to operate within. Default: cwd.
#ROOT_PATH=
# Sandbox mode used by the server. Same values as SANDBOX_ENV. Default: local.
#SANDBOX_ENV=local
# ----------------------------------------------------------------------------
# MCP (Model Context Protocol)
# ----------------------------------------------------------------------------
# Path to an MCP server config file (JSON). If unset, MCP is disabled.
#MCP_CONFIG_PATH=
# ----------------------------------------------------------------------------
# Model metadata override (MODEL_* — optional)
# ----------------------------------------------------------------------------
# These variables supply or override the active model's metadata — context
# window, pricing, capabilities, reasoning config, etc.
#
# modelInfo is also resolved from models.dev by MODEL name when online.
# Use MODEL_* to override or fill gaps for custom endpoints.
#
# All MODEL_* fields are optional; only set the ones you care about.
# Any one of them being set is enough to activate the override.
# Display name shown in the UI. Falls back to the MODEL value.
#MODEL_NAME=
# Optional base URL override stored in modelInfo (applied after connection resolution).
#MODEL_BASE_URL=
# Model string actually sent to the API. Defaults to the MODEL value.
#MODEL_API_NAME=
# Max input context window in tokens. Default if unset: 128000.
#MODEL_CONTEXT_WINDOW=128000
# Default max output tokens. Default if unset: 4096.
#MODEL_MAX_OUTPUT=4096
# Whether the model accepts image / multimodal input. true | false
#MODEL_MULTIMODAL=false
# Pricing in USD per 1M tokens (all optional; 0 when omitted).
#MODEL_PRICE_INPUT=0
#MODEL_PRICE_OUTPUT=0
#MODEL_PRICE_CACHE_WRITE=0
#MODEL_PRICE_CACHE_READ=0
# Comma-separated capability list. Known values:
# reasoning, vision, tool_calling, prompt_caching, streaming,
# json_output, computer_use
#MODEL_CAPABILITIES=tool_calling,streaming
# Reasoning model config (all optional).
#MODEL_REASONING_TAG=
#MODEL_REASONING_EFFORT=medium
#MODEL_REASONING_BUDGET=4096
# Mark this as the default model for its provider. true | false
#MODEL_IS_DEFAULT=false
# ----------------------------------------------------------------------------
# Node.js / build
# ----------------------------------------------------------------------------
# Usually set by the npm scripts; rarely needs manual configuration.
#NODE_ENV=production