Split a coding task across separate planner, executor, and reviewer AI models instead of relying on one.
Inspect exactly which model was used for each step of a task and why, along with its cost.
Preview which model the router would pick for a role before spending any API budget.
Track and audit past coding tasks through saved plan, decision, and cost files.
| ayushparkara/syntra | 1ncendium/aibuster | aaronmayeux/ha-hurricane-tracker | |
|---|---|---|---|
| Stars | 5 | 5 | 5 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | easy |
| Complexity | 3/5 | 3/5 | 2/5 |
| Audience | developer | ops devops | general |
Figures from each repo's GitHub metadata at analysis time.
Requires your own API keys configured in a separate providers.json file to run real tasks.
Syntra is a terminal based coordination tool for coding with more than one AI model at once, instead of trusting a single model with an entire task from start to finish. It splits the work into three roles: a planner model that breaks a task into steps, an executor model that carries out each step, and a reviewer model that checks the result. Each of these can be a different AI model, chosen based on which one is best suited for that particular job. Rather than keeping everything in one long chat log, Syntra saves structured files for each task, including the plan, the decisions made, any failures, a summary, and the cost. This makes it possible to inspect exactly what happened at each stage and why a particular model was chosen for a particular role. The main way to use Syntra is through a full screen terminal interface you open by simply typing syntra, which gives you panels for planning, execution, review, an activity trail, and the current model route, along with about 20 visual themes and a command palette. Everything available in the terminal interface also has a command line equivalent for people who prefer scripting or plain terminal use, such as running a single task with syntra run or asking the router which model it would pick without spending any money on an API call. Installation is straightforward using standard Python packaging tools like uv, pipx, or pip, or by cloning the repository directly and installing it in editable mode. Actually running a task requires configuring your own API keys in a separate provider configuration file kept outside the repository, since it should never be committed to version control. The model routing decisions are guided by a bundled catalog of AI models tagged with information like intelligence, speed, price, and which roles they suit best. This catalog is described as an approximate starting point rather than guaranteed live data, and users can inspect, override, or refresh it themselves. The project is explicitly in early public beta and the maintainers are looking for real world bug reports and feedback on where model routing goes wrong.
A terminal tool that coordinates multiple AI models for coding tasks, assigning separate models to plan, execute, and review each step.
Mainly Python. The stack also includes Python, CLI, TUI.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.