Our take
Quietno commits in the last 90 days- Activity
- 31
- Community
- 73
- Issues
- 21
- Pull requests
- 32
-3 stars in the last 30 days
- Slower to merge pull requests than 83% of the projects we track
Lagent is a flexible Python framework for developers who want to assemble and customize LLM agent workflows, though its measured project status is quiet and recent issue closure is absent.
Good fit if
- You want to build Python agents with custom message aggregation, memory, parsers, hooks, and tool execution.
- You need both synchronous and asynchronous workflow components and can keep each workflow’s components consistent.
- You are comfortable integrating model backends and supplying any required API credentials yourself.
Look elsewhere if
- You want a ready-to-use agent product rather than a framework to compose in Python.
- You require evidence of active commit-level development or recent issue resolution before adopting a dependency.
- Your project needs a specific platform or deployment package such as Docker, Compose, or Helm, which the provided project facts do not list.
All health signals
| Last commit | 2026-10-09 (5 days ago) |
|---|---|
| Commits, last 90 days | 0 |
| Releases, last 12 months | 1 (latest agentrl_rc0, 2026-05-19) |
| Contributors | 34 (top contributor: 28% of commits) |
| Issues closed, last 90 days | 0 |
| Pull requests merged, last 90 days | 4 (typically merged in 3 days) |
| Project age | 3 years |
Checked on 2026-10-15 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
2,281Forks
248Change since Nov 16, 2026: 0 stars·0 forks
Overview
InternLM’s Lagent is a lightweight Python framework for building LLM-based agents. Inspired by PyTorch’s design philosophy, it organizes agent workflows around components and message passing, aiming to make it easier to build and customize single-agent and multi-agent applications.
The framework provides agent messages, memory, response parsing, tool execution, and synchronous and asynchronous interfaces. It connects to model implementations such as vLLM and API-backed models, so it suits developers who want to orchestrate model calls and tools in Python rather than adopt a turnkey agent application.
Key Features
- Compose agents around message-based inputs and outputs with
AgentMessage. - Keep per-session conversation state in agent memory, and reset or inspect that state.
- Customize message aggregation, including adding few-shot examples.
- Parse structured model responses for tool calls or other output formats.
- Execute tools through
ActionExecutor, with hooks to adapt messages and results. - Use synchronous or asynchronous variants of components, including agents, model interfaces, and action executors.
- Build multi-agent workflows, such as writer-and-critic loops or research and charting pipelines.
Use Cases
- Python developers can prototype an LLM agent that calls a code interpreter and iterates on a problem.
- Teams can connect an agent to browser or other tools and customize how tool requests and results are passed through the workflow.
- Application developers can create separate agents for drafting and critique, or other cooperative tasks.
- Developers serving multiple concurrent requests can use asynchronous components with isolated session IDs.
What you need
Detected in the repository
- Python (from setup.py)
- A test suite and automated checks on GitHub Actions
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install from source:
git clone https://github.com/InternLM/lagent.git
cd lagent
pip install -e .
See the README and documentation for setup details and examples.
Alternatives
- OxyGent: OxyGent emphasizes modular components and observable multi-agent execution, while Lagent centers on a PyTorch-inspired design with synchronous and asynchronous workflows.
- ROMA: ROMA focuses on recursive planning for complex tasks, while Lagent provides a broader framework for composing agents, tools, memory, and message passing.
- autogen: AutoGen supports agents working independently or with people and is in maintenance mode; Lagent emphasizes synchronous and asynchronous workflows and agent composition.
- harness-sdk: Strands offers Python and TypeScript tools ranging from a ready-made harness to a customizable SDK, while Lagent uses a PyTorch-inspired Python framework design.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| lagent | Python | Apache-2.0 | 2.3k | Quiet |
| OxyGent | Python | – | 2.1k | Maintained |
| ROMA | Python | – | 5.2k | Quiet |
| autogen | Python | CC-BY-4.0 | 61.3k | Quiet |
| harness-sdk | Python | Apache-2.0 | 9k | Active |
Considerations
- GitHub health data classifies the project as quiet, with no commits recorded in the last 90 days. It also reports no issues closed in that period, although pull requests were merged and a release was published in the last year.
- The project has contributors across the repository, and measured contribution share is not concentrated in one person. This reduces the specific risk of depending on a single contributor, but does not remove the need to assess maintenance activity.
- The README demonstrates different model backends and tool integrations. Model serving resources, external services, and credentials depend on the backend and tools selected; the repository facts do not specify a universal hardware requirement.
- The project uses Python and has CI and tests. Docker, Compose, and Helm support are not listed in the supplied project facts.
- Apache-2.0 licensing permits broad use, subject to the license’s terms and obligations.
Found this useful?
Share it with someone who would like lagent.