VS Code extension

The Sagewai VS Code extension adds directive syntax highlighting, scaffold commands, and code snippets for agent development. It requires no configuration and activates automatically when you open a Python or Markdown file.

Installation

From VSIX (current)

cd vscode/
npm install
npm run compile
npx @vscode/vsce package

Then open the Command Palette in VS Code, run Extensions: Install from VSIX..., and select the generated .vsix file.

From source (development)

Open the vscode/ directory in VS Code
Press F5 to launch the Extension Development Host
The extension activates for Python and Markdown files automatically

VS Code Marketplace (coming soon)

ext install sagewai.sagewai

Directive syntax highlighting

The extension injects TextMate grammar patterns into Python and Markdown files. Directives are highlighted separately from regular Python strings, which helps you spot retrieval calls, agent delegations, and model overrides at a glance.

Supported directives:

Directive	Purpose	Example
`@context()`	Retrieve documents from Context Engine	`@context('quarterly earnings', scope='project', tags='finance')`
`@memory()`	Search stored facts	`@memory('customer preferences')`
`@agent:name()`	Delegate to another agent	`@agent:researcher('Find latest pricing data')`
`@wf:name()`	Invoke a saved workflow	`@wf:approval_pipeline('Review this contract')`
`/tool.name()`	Call a tool inline	`/tool.web_search('Sagewai documentation')`
`#model:name`	Override the model	`#model:gpt-4o`
`#budget:amount`	Set a budget limit	`#budget:1.00`
`@datetime`, `@date`, `@time`	Dynamic parameters	Resolved at prompt-processing time
`@user`, `@project`	Context parameters	Current user and project
`{{ expression }}`	Template expressions	`{{ context.search('query') }}`

Scaffold commands

Open the Command Palette (Cmd+Shift+P or Ctrl+Shift+P) to access these commands:

Sagewai: New Agent

Generates a complete agent file with proper imports, an example @tool decorated function, agent configuration, and an async usage example:

from sagewai import UniversalAgent, tool

@tool
async def search_web(query: str) -> str:
    """Search the web for information."""
    # TODO: implement
    return f"Results for: {query}"

agent = UniversalAgent(
    name="my-agent",
    model="gpt-4o",
    system_prompt="You are a helpful research assistant.",
    tools=[search_web],
)

# Usage: response = await agent.chat("Find the latest AI news")

Sagewai: New Workflow

Generates a multi-agent sequential pipeline:

from sagewai import UniversalAgent, SequentialAgent

researcher = UniversalAgent(
    name="researcher",
    model="gpt-4o",
    system_prompt="You research topics thoroughly.",
)

writer = UniversalAgent(
    name="writer",
    model="gpt-4o",
    system_prompt="You write clear, engaging content.",
)

pipeline = SequentialAgent(
    name="research-pipeline",
    agents=[researcher, writer],
)

# Usage: result = await pipeline.chat("Write an article about quantum computing")

Sagewai: Add Tool

Inserts a @tool decorated function at your cursor position. Tab stops let you fill in the function name, parameters, and docstring without touching the surrounding boilerplate.

Code snippets

Type a prefix and press Tab to expand:

Prefix	Expands to	Use case
`swagent`	Full agent with tool and config	Quick agent setup
`swtool`	`@tool` decorated function	Add a single tool
`swworkflow`	Sequential dual-agent pipeline	Multi-stage workflows
`swctx`	`@context('query', scope='project', tags='...')`	Context retrieval directive
`swmem`	`@memory('query')`	Memory search directive
`swag`	`@agent:name('task')`	Agent delegation directive

What the extension does for your workflow

Scaffold commands generate a complete, correctly structured file in a couple of seconds — correct imports, async patterns, @tool decorator, UniversalAgent config — rather than typing it from scratch.

Syntax highlighting makes directive strings visually distinct from regular Python strings. You can see at a glance which parts of a prompt are directives, which helps when editing multi-directive prompts.

Snippets with tab stops let you fill in the variable parts (name, system prompt, tool parameters) without hunting through boilerplate. Tab from placeholder to placeholder.

Works alongside AI assistants — Copilot, Cursor, and Claude Code can all see highlighted directives in your code. Snippets handle the scaffolding so your AI assistant focuses on the logic.

Configuration

No configuration is required. The extension activates automatically when you open any Python or Markdown file.

Requirements:

VS Code 1.85.0 or later
No runtime dependencies

Technical details

Activation events: onLanguage:python, onLanguage:markdown
Grammar injection: TextMate grammar patterns injected into Python and Markdown
Bundle size: Minimal — no runtime dependencies, compiles to a single JS file