# itayinbarr/little-coder

A harness optimized to smaller LLMs

Repository: https://github.com/itayinbarr/little-coder
Canonical: https://ross.abutalabs.com/products/little-coder
Homepage: https://itayinbarr.github.io/little-coder/
Language: TypeScript
License: Apache-2.0
License Family: permissive
Topics: ai-coding-assistant, benchmark, code-generation, coding-agent, coding-agents, local-llm, ollama, qwen, small-language-models, tool-use, aider-polygot, terminal-bench
Last push: 2026-08-22T11:41:16+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 99, longevity 10
- inputs: {"age_days": 144, "days_push": 11, "days_rel": 11, "gap_med": 0.0, "n_releases_24m": 45}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2504, forks 168 (observed 2026-08-28T04:06:57.018723+00:00)

## What it is
little-coder is a terminal coding agent built on the pi agent loop, tuned for small local language models like Qwen via Ollama or llama.cpp. It bundles pi extensions, skill files, and a Python benchmark harness, and ships as an npm-installable CLI that reads, edits, and runs code against your working tree offline.

## Use cases
- run a coding agent on a small local LLM offline
- benchmark small models on aider polyglot and terminal-bench
- edit code in my repo with a local qwen model
- use a coding assistant without sending code to the cloud
- test scaffold-model fit for coding agents
- serve a 35B MoE model on an 8GB laptop GPU for coding

## When to choose
- you want an offline, privacy-preserving coding agent powered by small local models
- you need to benchmark or evaluate small LLMs on coding tasks
- you have limited GPU/VRAM and want a tuned agent scaffold

## When to avoid
- you need frontier-model quality and prefer cloud APIs
- you rely on globally installed pi extensions, which don't load by default
- you need a GUI-based coding assistant

## Facets
- artifact type: cli-tool
- maturity: active
- function: agent-framework, llm-inference, developer-tools, cli, benchmarking
- domain: large-language-models, developer-tools
- platform: cli, windows
- tags: coding-agent, local-llm, ollama, small-language-models, tool-use, terminal-bench, aider-polyglot, qwen, benchmark-harness, pi-extensions, ai-agents, command-line, nodejs, macos, linux

## Member repositories
- itayinbarr/little-coder (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:57.018723+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:26:45.062829+00:00, confidence not recorded.
  - readme: https://github.com/itayinbarr/little-coder (fetched 2026-08-28T04:06:57.018723+00:00, sha b839545780ef)
  - homepage: https://itayinbarr.github.io/little-coder/ (fetched 2026-08-29T10:09:03.480579+00:00, sha 5118c29e7934)
  - registry_npm: https://registry.npmjs.org/little-coder (fetched 2026-08-29T10:09:03.490230+00:00, sha 2e1860813f72)
- Data as of 2026-08-30T08:39:29.467469+00:00.