Repolode
Submit
eastreams

eastreams/loong

A Rust base for building vertical AI agents, with a CLI, 42+ LLM providers, 25+ chat channels and a modular kernel.

View on GitHubWebsite
Stars
638
Forks
102
Open issues
48
Last commit
8 days ago

What is it?

Loong is an agent infrastructure project written in Rust, aimed at building AI agents for specific domains. It provides a single `loong` command-line entrypoint with an interactive terminal UI, one-shot queries, provider and channel setup, environment diagnostics, and self-update.

The project is organized as a 13-crate Rust workspace with a layered design. Providers, tools, channels, memory, and policy sit outside the core kernel, so they can be compiled in or composed as needed. Provider and channel configuration is written to a TOML file, with secrets read from environment variables. Feishu/Lark has a dedicated onboarding flow that binds a bot through the official registration API.

Loong supports approvals, policy, and audit within explicit runtime boundaries, and separates the assistant, gateway, and channels so they run independently. The README states that its configuration can be made compatible with existing setups from tools such as OpenClaw, Claude Code, and Codex.

Key features

  • 01Single `loong` CLI with an interactive TUI, one-shot `ask`, and `doctor` diagnostics.
  • 02Built-in support for 42+ model providers configured through TOML.
  • 0325+ channel integrations, including a guided Feishu/Lark bot setup.
  • 04Core kernel separated from providers, tools, channels, memory, and policy.
  • 05Approvals, policy, and audit operate within explicit runtime boundaries.

Use cases

  • Building a domain-specific AI agent on a Rust codebase.
  • Running an AI assistant that talks to a team through Feishu or Lark chats.
  • Switching between multiple LLM providers from one configuration file.
  • Studying a layered, modular agent architecture written in Rust.

Quick start

  1. 1

    Install a C linker and the Rust toolchain (rustup command is a piped installer, so run it from rustup.rs)

  2. 2

    Build and install the loong CLI from the repository source

    cargo install --path crates/daemon
  3. 3

    Run loong to open the main TUI; first-run setup starts there if config is missing

    loong

Things to know

  • Building from source requires a Rust toolchain and a C linker.
  • Most providers and channels need your own API keys or account credentials configured.

In collections

Open-source AI agent frameworks & appsCLI & terminal tools

Topics

#agent#rust#llm#agentic-ai#cli#feishu#multi-provider#agents#ai#assistant

FAQ

What is loong?
A Rust base for building vertical AI agents, with a CLI, 42+ LLM providers, 25+ chat channels and a modular kernel. Loong is an agent infrastructure project written in Rust, aimed at building AI agents for specific domains. It provides a single `loong` command-line entrypoint with an interactive terminal UI, one-shot queries, provider and channel setup, environment diagnostics, and self-update.
Is loong free to use? What license does it have?
loong is open source under the MIT license. Check the license terms before using it in commercial products.
Is loong still maintained?
Actively maintained. The last commit to loong was on Oct 14, 2026.
What can I use loong for?
Building a domain-specific AI agent on a Rust codebase. Running an AI assistant that talks to a team through Feishu or Lark chats. Switching between multiple LLM providers from one configuration file. Studying a layered, modular agent architecture written in Rust.
How do I install loong?
Install a C linker and the Rust toolchain (rustup command is a piped installer, so run it from rustup.rs) Build and install the loong CLI from the repository source: cargo install --path crates/daemon Run loong to open the main TUI; first-run setup starts there if config is missing: loong
Where can I get loong?
The source code and installation instructions are on GitHub: https://github.com/eastreams/loong

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects