AINL v1.1.0 — First Public Release

AINL is the open language for deterministic AI workflows. It compiles structured workflows to canonical graph IR and executes them through a deterministic runtime with adapter-based side effects, policy-gated execution, and multi-target emission.

Highlights

• Python 3.10+ official baseline, with CI coverage on 3.10 and 3.11
• Core test profile fully green — 403 tests, 0 failures
• MCP v1 server — a thin, stdio-only MCP server (ainl-mcp) exposing workflow compilation, validation, execution, capability discovery, and security introspection for Gemini CLI, Claude Code, Codex, and other MCP-compatible hosts
• HTTP runner service — POST /run with policy-gated execution, /capabilities discovery, health/metrics endpoints
• Security/operator surfaces — adapter privilege tiers, named security profiles, policy validator with forbidden_privilege_tiers, security report tooling
• Docs reorganized — intent-based information architecture with section READMEs and a root navigation hub

Start here

| Path | First step | |------|-----------| | CLI only | ainl-validate examples/hello.ainl --strict | | HTTP runner | ainl-runner-service then curl localhost:8770/capabilities | | MCP host | pip install -e ".[mcp]" && ainl-mcp |

Links

• Getting started
• Release notes
• External orchestration guide
• Security & threat model
• Conformance
• Contributing

Licensing

Apache 2.0 for core code. See LICENSE, LICENSE.docs, MODEL_LICENSE.md, and COMMERCIAL.md for full details.