Dev Tools
deepseek-cli
A command-line AI coding assistant powered by DeepSeek Coder models.
★ 319 starsMIT
View guide →A coding-agent harness optimized for smaller local LLMs — small models, serious coding work
little-coder is an open-source dev tools project by itayinbarr (itayinbarr/little-coder on GitHub). It has 2,655 GitHub stars and 183 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 18 September 2026.
Decide your model: a local llama.cpp/Ollama/LM Studio server for privacy and cost, or a cloud model through the same harness.
Install with the one-line script or npm — no clone or workspace install needed.
If local, start your model server (the README gives full llama.cpp, Ollama and LM Studio recipes) and set the provider's API-key env var (local servers ignore the value but need one present).
cd into your project, run little-coder, and give it a task — or start in plan mode (ctrl+q) for bigger work.
Extensions and skills are designed around scaffold–model fit, keeping cold-start context around 7k tokens so smaller models stay effective.
Research with sub-coders, 1–3 clarifying questions with suggested answers, an approved plan saved to .pi/approved-plan.md, then /implement starts a fresh session seeded with the plan.
f2 / /deep-research fans out read-only research sub-coders and returns one cited markdown report.
Isolated read-only child sessions research without cluttering the main context; long commands run as background jobs that wake the model on events.
Plan on a big model and implement on a small one with /plan-model and /action-model; sessions are named, resumable, and edits require reading the file first.
pi stays a plain dependency — no fork — and everything little-coder-specific lives under .pi/extensions, skills and benchmarks.
Clone little-coder only if you want to hack on it; installation is step 2.
git clone https://github.com/itayinbarr/little-coder.git
cd little-coderUse the one-line installer (Node.js 22.19+) or npm, as the README shows.
curl -fsSL https://raw.githubusercontent.com/itayinbarr/little-coder/main/install.sh | bash
# or: npm install -g little-coderStart your local model server first if using one, then run little-coder in the project directory.
cd ~/your-project
little-coder
# or name a model: little-coder --model llamacpp/qwen3.6-35b-a3bitayinbarr maintains it, building on the open pi agent substrate (pi.dev).
Yes, Apache-2.0 licensed. Local models are free to run; cloud models use your own provider keys.
No — its canonical setup is fully local (llama.cpp/Ollama/LM Studio). Local providers still need a dummy API-key env value set, as the README explains.
Those assume a frontier model. little-coder's whole design — small context, skill cards, plan/implement split — exists to make much smaller local models reliable at coding tasks.
Run the one-line install, serve a model with llama.cpp or Ollama, cd into a project and run little-coder.
A command-line AI coding assistant powered by DeepSeek Coder models.
Research into how agentic AI coding assistants work. Reconstructed prompt patterns, agent coordination, and security classification
A terminal coding agent built for DeepSeek models — reliable, cache-aware and cheap to run
A high-performance coding agent harness in Rust - fast TUI, swarms and every provider
Open-source AI coding assistant for VS Code. Code completion, chat, edits and reviews with local or hosted models. Your models, your infrastructure.
Alibaba's open-source terminal coding agent for Qwen models — a Gemini CLI-style workflow, free to run
The most-starred other Dev Tools projects in the LibreAI catalogue — the realistic shortlist if little-coder is not the right fit.
The open source coding agent
The model-definition framework behind modern AI — thousands of pretrained transformers for text, vision, and audio.
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
OpenAI's lightweight coding agent that runs locally in your terminal
Free and open — grab the source, star the repo, and start creating.