vtcode 0.15.7

A Rust-based terminal coding agent with modular architecture supporting multiple LLM providers
Documentation 
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

# VTCode Configuration
# Minimal configuration with only actively used settings

[agent]
# Provider to use: one of "gemini", "openai", "anthropic", "openrouter"
# API key environment variable is automatically inferred from provider
provider = "openrouter"
# Default model for single-agent mode
# Please also check router.models below if change default_model
default_model = "x-ai/grok-4-fast:free"
# Chat UI surface: auto | alternate | inline
ui_surface = "inline"

[router.models]
# Model routing tiers consumed by vtcode-core::router
simple = "x-ai/grok-4-fast:free"
standard = "x-ai/grok-4-fast:free"
complex = "x-ai/grok-4-fast:free"
codegen_heavy = "x-ai/grok-4-fast:free"
retrieval_heavy = "x-ai/grok-4-fast:free"


# Max conversation turns per session (prevents infinite loops)
max_conversation_turns = 150
# Reasoning effort for models that support it: "low" | "medium" | "high"
reasoning_effort = "medium"

[security]
# Runloop guardrails enforced inside session_setup
human_in_the_loop = true
require_write_tool_for_claims = true
auto_apply_detected_patches = false

[automation.full_auto]
# Full-auto safeguards consumed by ToolRegistry::enable_full_auto_mode
enabled = false
require_profile_ack = true
profile_path = "automation/full_auto_profile.toml"


[tools]
default_policy = "prompt"
max_tool_loops = 100

[tools.policies]
# File system tools safeguard workspace access
read_file = "allow"
list_dir = "allow"
grep_search = "allow"
run_terminal_cmd = "allow"
curl = "prompt"
create_file = "allow"
edit_file = "allow"
delete_file = "deny"

# Static analysis helpers available without prompts
# Code analysis tools
semantic_search = "allow"
tree_sitter_analyze = "allow"
# Build and format commands needed for Rust workflow

# Build and format commands needed for Rust workflow
# Build and test tools
cargo_check = "allow"
cargo_build = "prompt"
cargo_test = "allow"
cargo_clippy = "allow"
# Git inspection is safe; history mutation asks first
cargo_fmt = "allow"

# Git tools
git_status = "allow"
git_diff = "allow"
git_log = "allow"
# Powerful refactor/search requires confirmation
# Powerful refactor/search requires confirmation
srgn = "prompt"
git_push = "prompt"

# Code modification tools

[commands]
# Quick inspection commands the agent can run freely
allow_list = ["ls", "pwd", "git status", "git diff", "cargo check", "echo"]
# Hard-stop list for destructive or high-risk invocations
deny_list = ["rm -rf /", "rm -rf ~", "shutdown", "reboot", "sudo *", ":(){ :|:& };:"]
# Pattern-based permissions for common workflows
allow_glob = ["git *", "cargo *", "python -m *"]
deny_glob = ["rm *", "sudo *", "chmod *", "chown *", "kubectl *"]

[pty]
# PTY defaults used by vtcode-core::tools::registry::pty
enabled = true
default_rows = 24
default_cols = 80
max_sessions = 10
command_timeout_seconds = 300
stdout_tail_lines = 20

[ui]
# Controls inline tool output rendering in vtcode-core::ui
tool_output_mode = "compact"

[mcp]
# Local MCP transports; fine-grained allowlists live in .vtcode/tool-policy.json
# Local MCP clients executed via stdio transports
enabled = true

[[mcp.providers]]
# Official Model Context Protocol time server
name = "time"
enabled = true
command = "uvx"
args = ["mcp-server-time"]
max_concurrent_requests = 3

[[mcp.providers]]
# Upstash Context7 knowledge base over MCP
name = "context7"
enabled = true
command = "npx"
args = ["-y", "@upstash/context7-mcp@latest"]
max_concurrent_requests = 3

[[mcp.providers]]
# Anthropic sequential thinking planner via MCP
name = "sequential-thinking"
enabled = true
command = "npx"
args = ["-y", "@modelcontextprotocol/server-sequential-thinking"]
max_concurrent_requests = 3

# Example HTTP provider configuration
# Note: HTTP transport is partially implemented - basic connectivity testing is supported
# but full streamable HTTP MCP server support requires additional implementation
# [[mcp.providers]]
# name = "example-http"
# enabled = false
# command = "curl"  # Not used for HTTP transport
# args = []         # Not used for HTTP transport
# [mcp.providers.transport.http]
# endpoint = "http://localhost:3000/mcp"
# api_key_env = "HTTP_API_KEY"
# protocol_version = "2024-11-05"
# headers = { "Content-Type" = "application/json", "User-Agent" = "vtcode-mcp-client" }
# max_concurrent_requests = 2