A purpose-built DSL for defining, orchestrating, and shipping AI agent workflows. Readable syntax, multi-model support, runs anywhere.
# Define an agent that triages GitHub issues
agent IssueTriager {
model: "claude-sonnet-4"
budget: 50 # max LLM calls before forcing a decision
skills: [github, notify]
on issue_opened(issue: Issue) {
let analysis = analyze(issue.body)
match analysis.priority {
"critical" → github.label(issue, "P0")
→ notify.slack("#oncall", issue)
"high" → github.label(issue, "P1")
_ → github.label(issue, "needs-triage")
}
}
}Why llanguage
Everything you need to build reliable, observable AI agent workflows without the boilerplate.
A DSL designed specifically for agent orchestration — not a library bolted onto an existing language. Reads like intent, not plumbing.
Connect to any API, MCP server, or service with a simple plugin definition. GitHub, Slack, AWS, and more — built-in and community-contributed.
Target any LLM provider — Anthropic, OpenAI, local models — and switch without rewriting your agent logic.
Write, validate, and run agents in your browser. Live parse errors, spec conformance checking, and instant feedback.
Set hard limits on LLM calls, token spend, and tool invocations per agent run — so a runaway loop never surprises your bill.
A single compiled binary. Run on your laptop, in CI, as a Lambda, or inside a Docker container — no runtime dependencies.
How it works
Write a .ll file. Declare your model, skills, and workflow logic in a clean, readable syntax.
Run ll validate for instant spec conformance. Use the visual editor for live feedback.
Execute with ll run agent.ll. Pass inputs via --input key=value flags or compose agents with imports.
Embed the ll binary in any CI/CD pipeline, container, or serverless function.
LL ships with a growing library of community plugins — each one built by the plugin-builder agent itself, documented, and published automatically.
Search repos, read files, open issues and pull requests, and react to webhooks — all from LL agents.
officialPost messages, create channels, and listen for Slack events inside any LL workflow.
official
The plugin-builder.ll agent researches gaps, designs a plugin spec,
writes the implementation, updates the registry, and opens a PR — hands-free.
Browse the full registry → plugins/registry.json · Build your own with examples/plugin-builder.ll