Gantrix
Gantrix is an open-source AI agent orchestrator that plans a goal as dependent steps, runs ready steps in parallel, and has a judge check each result. Agents use skills and can retry rejected work or replan when a step keeps failing.
About Gantrix
Gantrix is an open-source AI agent orchestrator that runs on your own machine. According to its project page, you give it a goal and a team of agents, and it splits the goal into dependent steps, runs ready steps in parallel, and has a judge review each result. Rejected work is retried with feedback, and the plan can change if a step keeps failing.
The page says it uses local models through Ollama by default but can also connect to OpenAI, Anthropic or other compatible servers. Agents get tools through skill folders, and risky tool calls can require approval. The page credits Per Diem and lists an MIT license. I found no independent reviews or press coverage, so these details come only from the project's own page.
Key features
- Dependent step planning
- Parallel step execution
- Judge checks each result
- Skills in Claude skill format
- Optional approval for risky tool calls
- Ollama by default
Frequently asked questions about Gantrix
- What is Gantrix?
- Gantrix is an open-source AI agent orchestrator that plans a goal as dependent steps, runs ready steps in parallel, and has a judge check each result. Agents use skills and can retry rejected work or replan when a step keeps failing. It is listed on botshub under coding agents.
- What can Gantrix do?
- Key features include dependent step planning, parallel step execution, judge checks each result, skills in Claude skill format and optional approval for risky tool calls.
- What are alternatives to Gantrix?
- Other coding agents on botshub include Alondra, Brix, DevReady and Keploy. See all alternatives to Gantrix.