PS-LANG™
Privacy-First Scripting Language for Multi-Agent Context Control
Control what AI agents see in your workflows with zone-based syntax.
What is PS-LANG?
PS-LANG is a flexible, platform-agnostic syntax for controlling what AI agents see in multi-agent workflows. Use zones to mark content as private, pass-through, or active workspace.
Designed to evolve with you:
- Override and extend the syntax for your needs
- Platform/agent/model agnostic approach
- Works with Claude, GPT, Cursor, Copilot, and custom agents
- Community-driven development
Perfect For
Multi-Agent Pipelines
Research → Writing → Review workflows with controlled context handoffs between agents.
Context Engineering
Control exactly what each agent sees in your multi-step workflows.
Benchmarking
Create clean test environments with isolated context for accurate benchmarks.
Privacy-First AI
Keep sensitive notes private while collaborating with AI assistants.
Why Use PS-LANG?
You don't even need to install it. As someone who writes prompts daily (Claude Code, Cursor, ChatGPT, etc.), PS-LANG just adds structure.
// Before: Unstructured prompt
"Hey Claude, build a data pipeline. Don't show the next agent my rough notes about the schema."
// After: Structured with PS-LANG
<@. Build ETL pipeline for user analytics @.>
<. Private: still deciding between Convex or Supabase realtime .>
<#. Next agent: uses daily batch processing, 24hr retention #.>
What this gives you:
Context Control
Decide what each agent sees
Cleaner Handoffs
Next agent gets only what they need
Better Logs
Your work becomes structured, searchable
Future-Proof
Syntax works today, unlocks encryption later
Zone Syntax
PS-LANG zones are patterns, not limits. Start with these 7 recommended zones, then invent your own.
<. text .>
Only current agent sees
<#. text #.>
Documentation for next agent
<@. text @.>
Current agent's work area
<~. text ~.>
AI-generated metadata
<$. text $.>
Business strategy, pricing
<?. text ?.>
Open questions
<.bm text .bm>
Metrics or references
Infinitely extensible: Create any zone that makes sense for your workflow. The pattern is <symbol. content symbol.>
v0.1.0-alpha.1
PS-LANG is in active alpha testing. We're gathering feedback on zone syntax, CLI tools, and real-world usage patterns.
Help us test:
- Is the zone syntax clear and intuitive?
- Are 7 zones enough, or do you need custom zones?
- Does it solve your multi-agent workflow problems?
- Any performance issues or integration friction?
Roadmap
Alpha (Current)
Basic zone syntax, CLI tools, VS Code themes, and community feedback collection.
Beta
.psl file format, secrets management, and enhanced preprocessing tools.
v1.0
Parser library, multi-agent framework integration, and official ChatGPT support.
.PSL File Format
.psl stands for Prompt Script Language — a structured file format for organizing multi-agent workflows, context handoffs, and AI collaboration sessions.
Think of .psl files as executable prompts with zones:
- Store multi-step workflows with zone-based context control
- Version control your prompt engineering work
- Share reproducible AI workflows with your team
- Keep sensitive notes private with agent-blind zones
Coming in Beta:
File Templates
Pre-built .psl templates for common workflows
Secrets Management
Encrypted zones for API keys and credentials
Preprocessing Tools
CLI commands to validate and transform .psl files
IDE Support
Syntax highlighting for VS Code and other editors
Get Started
Start using PS-LANG in your prompts today, or install the CLI for advanced features.
PS-LANG™ is a trademark of Vummo Labs · MIT License