openmake

openmake_llm: Self-Hosted AI Workspace with vLLM

OpenMake is a self-hosted AI workspace and agent runtime built around vLLM and LiteLLM. It coordinates models, agents, MCP tools, and research in one workspace for people who want to run AI work on infrastructure they control.

ActiveTypeScriptApache-2.0Self HostedLLMAI Agents
Stars
44
Forks
8
Stars, last 30 days
+1
Last commit
today
Contributors
5

Our take

Activeregular commits and releases

Health score

55/100

How it is scored
Activity
90
Community
22
Issues
21
Pull requests
86

+1 stars in the last 30 days

  • Merges more pull requests than 100% of the projects we track
  • Releases more often than 93% of the projects we track

OpenMake is a broad, actively released self-hosted workspace for people who want agent orchestration and model choice, though its young age and highly concentrated contribution history warrant a measured rollout.

Good fit if

  • You want one workspace to coordinate vLLM-backed models, agents, research, and MCP tools.
  • You can operate a Node.js application and its data services, and want to run it on Linux or macOS, or Windows through WSL2.
  • You value an Apache-2.0 project with frequent recent commits and releases, and can tolerate a project whose contributions are concentrated among a small number of people.

Look elsewhere if

  • You only need a simple chat interface and do not need agent tasks, model routing, or tools.
  • Your deployment must avoid Node.js 24 or cannot accommodate the services and operational setup described in the self-hosting guide.
  • You need a project with broadly distributed development ownership; most of the code comes from one contributor.
All health signals
Last commit2026-11-16 (today)
Commits, last 90 days100+
Releases, last 12 months100+ (latest v1.139.2, 2026-11-16)
Contributors5 (top contributor: 94% of commits)
Issues closed, last 90 days0+
Pull requests merged, last 90 days100+ (typically merged in 0 days)
Project age10 months

Checked on 2026-11-17 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

44

Forks

8

Change since Nov 16, 2026: 0 stars·0 forks

Overview

OpenMake is a self-hosted AI workspace and agent runtime built around vLLM and LiteLLM. It coordinates models, agents, MCP tools, and research in one place, addressing the gap between simply chatting with a model and using models and tools to complete multi-step work. The project supports Linux and macOS, with Windows use through WSL2.

It is aimed at people and teams who want to choose their own models and control where AI work runs. A local vLLM server is one option; OpenMake can also use hosted, OpenAI-compatible endpoints. The application brings model routing, agent tasks, deep research, and tool execution into a single workspace.

Key Features

  • Routes work to models by role and capability through vLLM and the LiteLLM gateway, with support for local and OpenAI-compatible endpoints.
  • Plans multi-step requests, runs capability tasks in parallel where possible, and combines their results into an answer.
  • Runs persistent agent tasks with pause, resume, approvals, checkpoints, and recovery after server restarts.
  • Provides deep research that searches sources and produces cited reports, with PDF and DOCX export.
  • Connects MCP servers, built-in tools, skills, and extensions, with an approvals inbox for items that need review.
  • Runs agent code and MCP servers in Docker sandboxes; local execution is also available through the Companion app or Code CLI.
  • Offers an OpenAI-compatible API, a Discord gateway bot, and model comparison through OpenMake Bench.

Use Cases

  • A homelab operator can host the workspace and connect it to local models served by vLLM or an existing OpenAI-compatible endpoint.
  • A researcher can delegate a topic to Deep Research and use the resulting cited report as a starting point for further work.
  • A developer can give an agent a goal involving files or code, with sandboxed execution and approval gates for risky actions.
  • A team using multiple model capabilities can route text, vision, image, or speech tasks to separately assigned models.

What you need

Detected in the repository

  • Node.js >=24 <25 (from package.json)
  • 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

curl -fsSL https://raw.githubusercontent.com/openmake/openmake_llm/main/install.sh | bash

The installer sets up the application and, on a fresh machine, can configure the wider stack. See the README and self-hosting guide for install options and model configuration.

Alternatives

  • deer-flow: DeerFlow focuses on long-running research, coding, and content workflows with subagents and sandboxes, rather than a broader model-and-agent workspace.
  • swarmclaw: SwarmClaw is a TypeScript multi-agent runtime centered on memory, schedules, and delegation, rather than coordinating models and research in one workspace.
  • AgentRun: AgentRun manages agent revisions and durable runs through a workflow DSL, rather than providing a broader workspace for models, tools, and research.
ProjectLanguageLicenseStarsStatus
openmake_llmTypeScriptApache-2.044Active
deer-flowPythonMIT83.4kActive
swarmclawTypeScriptMIT697Quiet
AgentRunTypeScriptApache-2.0423New

Considerations

  • This is a young project. Its measured activity is strong, with regular recent commits and releases, but development is highly concentrated: most contributions come from one person. The measured data also reports no issues closed in the recent period, so assess your support expectations rather than inferring responsiveness from release activity alone.
  • The application requires Node.js 24. The README describes PostgreSQL and Redis in its stack, and Docker-based sandboxes for agent and MCP execution; sandbox features are opt-in. Local vLLM deployments also require suitable inference infrastructure, while hosted compatible endpoints are an alternative.
  • Several substantial capabilities are explicitly planned rather than shipped, including an execution graph, declarative policy engine, and scoped memory. Confirm release status before depending on roadmap items.
  • Apache-2.0 is the project license. Review its terms for your intended use and distribution.

Found this useful?

Share it with someone who would like openmake_llm.

Comparisons