Build LLM applications
in pure Markdown
A declarative framework for building modular AI applications. Define tools and instructions in .md files.
---
tools:
- [psql, -d, $DB, -c, { regex: "^SELECT\b.*" }]
---
# Instructions:
- Run `psql` for read-only PostgreSQL queries
SIMPLE SYNTAX
Declarative by design
Define your AI agents in plain Markdown files. No complex configurations, no boilerplate—just clear, readable definitions that anyone on your team can understand and modify.
See the syntax Feature illustration
SANDBOXED EXECUTION
Tools without the complexity
Give your agents access to shell commands, APIs, and databases with built-in security constraints. Define regex patterns for safe execution without writing wrapper code.
Explore tools Feature illustration
MODULAR ARCHITECTURE
Composable agents
Build complex workflows by composing simple agents. Import capabilities from other Markdown files. Version control your AI logic alongside your code.
Learn composition Feature illustration
FULL TRANSPARENCY
Observable by default
Every decision, every tool call, every response—fully logged and traceable. Debug your agents with complete visibility into their reasoning process.
View observability Feature illustration