BAML

What it does

BAML is a typed, open-source programming language and toolchain for writing AI functions and agent workflows. Developers declare a function's input and output types, call a model through it, and get structured data back, with schema-aligned parsing repairing malformed model output against the declared types.

The runtime traces and profiles every function locally, supports streaming, batching and workflow primitives such as concurrency, retries and cancellation, and includes a test framework that can grade flaky AI functions over many runs. It runs standalone on macOS, Linux and Windows, or inside existing projects through generated bridges to Python, TypeScript, Java, Go, Rust and other languages.

It suits developers building LLM-powered features who want typed structured output and tracing without rewriting an existing codebase. Teams should expect to write or port some libraries themselves, since BAML ships a standard library rather than a package ecosystem.

What it can do

• Typed AI functions that declare input and output types and return structured data from model calls.
• Schema-aligned parsing that repairs malformed model output against declared types.
• Always-on tracing and profiling of every function, with traces readable by agents.
• Streaming, batching, and websocket support for AI functions across providers.
• Built-in test generation from datasets or production traces, with grading of flaky AI functions over many runs.
• Concurrency and workflow primitives including retries and cancellation.

Works with

• Python
• TypeScript (Node and Web)
• Java
• C#/.NET
• C++
• Go
• Rust
• Kotlin
• Swift
• PHP and Ruby (described as coming soon)

Before you choose

• BAML has no npm-style package ecosystem; additional libraries may need to be written, ported or passed across the bridge.
• A paid cloud with observability, team controls and governance is announced for later this year, with no published pricing.
• PHP and Ruby bridges are described as coming soon rather than available.

Listed under

• Coding
• MCP and agent tools
• Agents
• Developer tools
• LLM
• Observability
• Open source
• Programming language

Get started

Install via Homebrew, curl, Arch, or PowerShell as listed on the quickstart.

1. Install the BAML toolchain using Homebrew, curl, Arch (yay), or PowerShell as listed on the quickstart.
2. Run baml init and baml agent install in the project, then baml run main to verify.
3. For an existing app, run baml bridge add <target> once, install the printed host package, and run baml bridge generate when .baml files change.

BAML is free and open source (Apache-2) and does not call Boundary's servers. Cloud pricing has not been published; verify cost with Boundary before adopting the cloud offering.

BAML is free to use. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Basically a Made-up Language ↗boundaryml.com
• Quickstart | BAML ↗boundaryml.com
• Pricing | BAML ↗boundaryml.com

Company website

boundaryml.com: product, pricing and policy pages

Hacker News

Launch posts and discussion

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

manual_review

Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.