Tools Overview¶
Complete documentation for all hanzo-tools-* packages.
Architecture: Tools follow HIP-0300 - the Unified MCP Tools Architecture with orthogonal operators and effect tracking.
Installation¶
# All tools (recommended)
pip install hanzo-mcp[tools-all]
# Core development tools
pip install hanzo-mcp[tools-dev]
# Individual packages
pip install hanzo-tools-shell hanzo-tools-browser
Tool Categories¶
HIP-0300 Core Operators¶
Primary operators organized by orthogonal axes:
| Tool | Axis | Actions | Description |
|---|---|---|---|
| fs | Bytes + Paths | read, write, edit, search, patch, tree | Filesystem operations |
| id | Identity | hash, uri, ref, verify | Content-addressable identity |
| code | Symbols + Structure | parse, transform, summarize | Code analysis and transformation |
| proc | Execution | run, bg, signal, wait | Process execution |
| vcs | History + Diffs | status, diff, commit, log | Version control |
| test | Validation | check, build, test | Validation loops |
| net | Network | search, fetch, download, crawl | Network operations |
| plan | Orchestration | intent, route, compose | Intent routing |
Control Surfaces¶
| Tool | Surface | Description |
|---|---|---|
| browser | Web DOM | Playwright automation (70+ actions) |
| computer | OS Desktop | Mac automation via pyautogui |
Extended Operators¶
| Tool | Domain | Description |
|---|---|---|
| lsp | Semantic Stream | Language server protocol (diagnostics, code_actions) |
| memory | Knowledge | Persistent memory and knowledge bases |
| todo | Task Tracking | Task management |
| reasoning | Cognition | Structured thinking (think, critic) |
| agent | Multi-Agent | Agent orchestration (run, list, status) |
| llm | LLM Interface | Unified LLM interface (llm, consensus) |
Infrastructure¶
| Package | Tools | Description |
|---|---|---|
| Core | Base classes | BaseTool, IdTool, ToolRegistry |
| Config | 2 | Configuration and mode management |
| Database | 8 | SQL and graph database operations |
| Vector | 3 | Semantic search with embeddings |
| Refactor | 1 | Code refactoring (rename, extract, inline) |
| Jupyter | 1 | Notebook read/edit/execute |
| Editor | 3 | Neovim integration |
| MCP | 4 | MCP server management |
Quick Reference¶
Most Used Tools¶
# File operations
read(file_path="/path/to/file")
write(file_path="/path/to/file", content="...")
edit(file_path="/path/to/file", old_string="old", new_string="new")
# Command execution
cmd("ls -la")
cmd(["npm install", "npm build"], parallel=True)
# Search
search(pattern="TODO", path=".")
ast(pattern="def test_", path="/tests")
# Reasoning
think(thought="Analyzing the problem...")
critic(analysis="Code review findings...")
# Browser
browser(action="navigate", url="https://example.com")
browser(action="click", selector="button")
Tool Discovery¶
from importlib.metadata import entry_points
# Discover all available tools
for ep in entry_points(group="hanzo.tools"):
tools = ep.load()
print(f"{ep.name}: {[t.name for t in tools]}")
Architecture¶
All tools follow a consistent pattern:
from hanzo_tools.core import BaseTool, ToolContext
class MyTool(BaseTool):
name = "my_tool"
@property
def description(self) -> str:
return "Tool description"
async def call(self, ctx: ToolContext, **params) -> str:
# Implementation
return result
See Also¶
- Core Libraries - Supporting packages (async, consensus, network)
- MCP Reference - Server documentation
- Agent Framework - Multi-agent SDK