A field guide to escaping GitHub Copilot's cloud
A living survey of local LLM tools for code completion, project generation, and shell assistance—curated because the README is literally generated from YAML.

What it does
This repo is a hand-maintained directory of open-source and local alternatives to GitHub Copilot. It catalogs editor extensions, project-generating tools, chat interfaces, models, and datasets—then auto-generates its own README from a YAML file. Think of it as a periodically updated map of a territory that changes weekly.
The interesting bit
The project eats its own dogfood in a small way: the README you read is compiled by a Python script from structured data, so the list itself is version-controllable and community-editable via pull requests. The maintainer also built one of the listed tools (gptme), which keeps the curation grounded in actual use.
Key highlights
- Covers 15+ editor extensions including Tabby, Continue.dev, twinny, and localpilot
- Surveys project-generation tools from gpt-engineer to aider to Refact.AI
- Lists chat interfaces with shell access, including open-interpreter and the maintainer’s own gptme
- Tracks relevant models: CodeLlama, Qwen 2.5 Coder, Deepseek R1, and others
- Includes a concrete “suggested setup” using LM Studio + Continue.dev with specific model picks for 2025
- README is generated from
data.yamlviagenerate_readme.py
Caveats
- The models section openly admits it’s “doomed to be out of date” due to release velocity
- Some listed tools are marked stale or unmaintained (Fauxpilot, terminal-x)
- Output quality of local copilots is explicitly noted as not yet matching cloud-based services
Verdict
Worth bookmarking if you’re actively building or choosing a local coding assistant stack. Skip it if you want a turnkey product—this is a reference, not a tool.