kimi-code/.env.example at main · fusengine/kimi-code · 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
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
# Fusengine Kimi Plugins - Environment Variables
# Copy to $KIMI_CODE_HOME/.env (default ~/.kimi-code/.env) and fill in your API keys.
# Kimi Code does NOT auto-load .env — the installer reads it when merging
# MCP servers into ~/.kimi-code/mcp.json, and the harness reads it at runtime.

# ============================================
# REQUIRED - Core MCP servers
# ============================================

# Context7 MCP - Up-to-date documentation lookup
# Get your key at: https://context7.com
CONTEXT7_API_KEY=ctx7sk-your-key-here

# Exa MCP - AI-powered web search and code context
# Get your key at: https://exa.ai
EXA_API_KEY=your-exa-key-here

# 21st.dev Magic MCP - AI UI component generation
# Get your key at: https://21st.dev/mcp (old MAGIC_API_KEY keys were reset —
# the 21st MCP shim reads API_KEY / TWENTY_FIRST_API_KEY / API_KEY_21ST only)
TWENTY_FIRST_API_KEY=your-21st-key-here

# Gemini Design MCP - AI frontend generation (Tailwind, shadcn)
# Get your key at: https://gemini-design-mcp.com (free: 20K tokens/month)
GEMINI_DESIGN_API_KEY=your-gemini-design-key-here

# ============================================
# OPTIONAL - Additional MCP servers
# ============================================

# GitHub MCP - Repos, PRs, issues, commits
# Get your token at: https://github.com/settings/tokens
# Required scopes: repo, read:org
GITHUB_TOKEN=ghp_your-token-here

# Brave Search MCP - Private web search
# Get your key at: https://brave.com/search/api
BRAVE_API_KEY=your-brave-key-here

# Serper MCP - Google search (SERP) and webpage scraping
# Get your key at: https://serper.dev/api-key
SERPER_API_KEY=your-serper-key-here

# Supabase MCP - Database, auth, storage
# Get your token at: https://supabase.com/dashboard/account/tokens
SUPABASE_ACCESS_TOKEN=sbp_your-token-here

# Notion MCP - Pages, databases, knowledge base
# Get your token at: https://www.notion.so/my-integrations
NOTION_TOKEN=secret_your-token-here

# Slack MCP - Messages, channels, workspace
# Get your token at: https://api.slack.com/apps
SLACK_TOKEN=xoxb-your-token-here

# Stripe MCP - Payments, invoices, customers
# Get your key at: https://dashboard.stripe.com/apikeys
STRIPE_SECRET_KEY=sk_test_your-key-here

# Sentry MCP - Error tracking and debugging
# Get your token at: https://sentry.io/settings/auth-tokens
SENTRY_AUTH_TOKEN=sntrys_your-token-here

# Replicate MCP - Run 1000+ AI models
# Get your token at: https://replicate.com/account/api-tokens
REPLICATE_API_TOKEN=r8_your-token-here

# ============================================
# DATABASE - For postgres MCP server
# ============================================

# PostgreSQL connection string (optional)
# DATABASE_URL=postgresql://user:password@localhost:5432/dbname

# ============================================
# OPTIONAL - Plugin behavior tuning
# ============================================

# APEX/SOLID enforcement TTL, in SECONDS (core-guards + ai-pilot hooks).
# Freshness window for the "explore+research called" / "SOLID ref read" checks
# and the statusline OAuth refresh cadence.
# UNSET = 120 (2 minutes) — this is the default if you leave it commented.
# Recommended values: 120 (2min) | 240 (4min) | 480 (8min) | 600 (10min).
# Note: the statusline daemon reads this at startup — restart it to apply a change.
# FUSE_ENFORCE_TTL_SEC=120

# SOLID max lines per file (core-guards + ai-pilot file-size hooks).
# Files over this limit are blocked (Write) / warned (Edit, PostToolUse, task validation).
# The split-headroom advisory = this value minus 10 (default 90).
# UNSET = 100 — this is the default if you leave it commented.
# Recommended values: 100 (strict SOLID) | 120 | 150 (Swift-style) | 200.
# FUSE_SOLID_MAX_LINES=100

# SOLID reference skills root for the @fusengine/harness solid-read gate.
# Set by the installer in settings.env → a path-delimiter list of every
# plugins/*/skills/solid-*/references dir (solid-generic/react/nextjs/php/swift/
# go/java/ruby/rust/csharp/python/astro), routed per file by the harness. Strong
# routing (applies-to globs) currently exists for TS/JS (solid-generic); other
# languages route by keywords. Override only to point at custom SOLID refs.
# FUSE_HARNESS_REFS=/abs/path/to/solid/skills

# Gemini design gates (@fusengine/harness `design` scope) — OPT-IN, off by default.
# When unset/0/false the Gemini-design enforcement (phase-order gates for the
# design-expert pipeline) is fully disabled. Set to 1/true to enable it.
# UNSET = disabled.
# FUSE_DESIGN_GEMINI=0

# Enforce Gemini Design MCP for UI writes (@fusengine/harness `gemini-mcp` gate).
# OPT-IN: off unless 1/true. When on, a hand-written .tsx/.jsx/.vue/.svelte write
# is blocked until a mcp__gemini-design__* call is made this session. UNSET = disabled.
# FUSE_ENFORCE_GEMINI_MCP=0

# Autonomous "Ralph" mode (@fusengine/harness git-command gate).
# OPT-IN: off unless 1/true. When on, the SAFE git commands skip the confirmation
# prompt (for unattended/agent loops). UNSET = disabled — safe commands still confirm.
# RALPH_MODE=0

# ============================================
# OPTIONAL - Harness cache & TTL tuning (settings.env; sane defaults if unset)
# ============================================

# Minimum minutes between two LESSON.md write-reminders (fuse-lessons Stop hook).
# UNSET = 5. Writing to MEMORY/LESSON.md resets the throttle regardless.
# FUSE_LESSONS_THROTTLE_MIN=5

# Cached MCP-doc freshness, in SECONDS (harness doc-consultation cache).
# UNSET = 172800 (48h). Lower it to re-consult docs more often.
# FUSE_MCP_TTL_SEC=172800

# WebFetch cache freshness, in SECONDS (pages go stale faster than docs).
# UNSET = 86400 (24h).
# FUSE_WEBFETCH_TTL_SEC=86400

# Subagent context-cache TTL, in MINUTES (SubagentStart injection cache).
# UNSET = 30.
# FUSENGINE_CACHE_TTL_MIN=30

# Marketplaces the harness scans for SOLID refs / skills, comma-separated.
# UNSET = fusengine-plugins.
# FUSE_HARNESS_MARKETPLACES=fusengine-plugins

# ============================================
# OPTIONAL - Neural Memory (Graphiti) endpoint
# ============================================

# Host + port of the Neural Memory (Graphiti) service: http://<host>:<port>.
# UNSET = localhost:8000.
# NEURAL_MEMORY_HOST=localhost
# GRAPHITI_PORT=8000