Doorman11991/smallcode

AI coding agent optimized for small LLMs. 87% benchmark with 4B-active model.

About the project

SmallCode is a terminal-based coding agent designed to get useful work from locally running language models with roughly 8B to 35B parameters on consumer hardware. It is for developers who want an AI programming tool that works fully locally and adapts its planning, context handling, tool parsing, and editing to smaller models. The catalogue entries are agents for SmallCode.

2.0kStars on the repository
11Mods indexed here, across every type
2mo agoLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown

code-engineer

01

Doorman11991/smallcode

Agent Claude Code

Primary implementer for any coding task — implementation, refactoring, debugging, code review.

not rated 2.0k +2 2mo ago A 22 tokens original MIT

critic

02

Doorman11991/smallcode

Agent Claude Code

Ruthless post-implementation verifier — rejects work that doesn't meet spec. Read-only except running checks.

not rated 2.0k +2 2mo ago A 23 tokens original MIT

debugger

03

Doorman11991/smallcode

Agent Claude Code

Systematic root-cause diagnosis — reproduce, hypothesize, test, fix, verify.

not rated 2.0k +2 2mo ago A 21 tokens original MIT

documenter

04

Doorman11991/smallcode

Agent Claude Code

Writes and updates docs — READMEs, inline comments, usage examples — matching the project's existing style.

not rated 2.0k +2 2mo ago A 24 tokens original MIT

general-purpose

05

Doorman11991/smallcode

Agent Claude Code

Catch-all agent for open-ended, multi-step tasks — research, content authoring, and text transformation (e.g. remastering/rewriting a section per a prompt or spec). Use when no more specific agent fits.

not rated 2.0k +2 2mo ago A 49 tokens original MIT

librarian

06

Doorman11991/smallcode

Agent Claude Code

External docs and library best-practices lookup — official references, real-world examples, GitHub repo discovery.

not rated 2.0k +2 2mo ago A 26 tokens original MIT

oracle

07

Doorman11991/smallcode

Agent Claude Code

Read-only architecture advisor — deep analysis, hard debugging, security and performance consulting.

not rated 2.0k +2 2mo ago A 18 tokens original MIT

planner

08

Doorman11991/smallcode

Agent Claude Code

Read-only; researches the codebase and produces a numbered, verifiable step plan before implementation.

not rated 2.0k +2 2mo ago A 21 tokens original MIT

qa-tester

09

Doorman11991/smallcode

Agent Claude Code

Writes tests, builds test suites, and discovers edge cases across unit, integration, and E2E levels.

not rated 2.0k +2 2mo ago A 26 tokens original MIT

red-team

10

Doorman11991/smallcode

Agent Claude Code

Adversarial security reviewer — find vulnerabilities, injection risks, exposed secrets, and failure modes. Read-only probing.

not rated 2.0k +2 2mo ago A 26 tokens original MIT

scout

11

Doorman11991/smallcode

Agent Claude Code

Fast read-only codebase recon — find files, patterns, functions, and entry points.

not rated 2.0k +2 2mo ago A 21 tokens original MIT

At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: