Hottest heavyweights

the big names on the move
01
morluto/rea
+8890 ★/day↘cooling

REA hands reverse-engineering tools — Hopper, Ghidra, process capture, JS reconstruction — to coding agents as one consistent MCP workflow.

★ 385.3k TypeScript Coding Assistants · explained Feature
02
Niko1221/Strata
+4437 ★/day↗accelerating

A 125-billion-parameter model that normally needs a server, packed onto one gaming PC — NVIDIA card plus RAM plus SSD — writing answers faster than you can read.

★ 32.9k C++ Inference · Serving · explained Feature
03
mattpocock/skills
+3379 ★/day↗accelerating

Composable prompt recipes that force coding agents to align, test, and architect instead of blindly generating code.

★ 301.8k Shell Coding Assistants · explained Feature
05
lexmount/moli
+2845 ★/day↗accelerating

Moli is a standalone Rust headless browser that avoids continuous rendering by computing layout and paint only when an operation explicitly demands geometry or pixels.

★ 20.7k Rust Agents · explained Feature
07
deepseek-ai/DeepInfer
+2296 ★/day→steady

DeepInfer is the inference framework DeepSeek actually runs at scale, now released for offline use — kernels, parallelism tricks and all.

★ 22.5k Python Inference · Serving · explained Feature
09
debpalash/VoiceStudio
+1647 ★/day↘cooling

OmniVoice Studio bundles voice cloning, dubbing, dictation, and TTS into a desktop app that keeps all audio processing off the internet and away from API keys.

★ 87k Python Image · Video · Audio · explained Feature
10
NandhaKishorM/laya
+1600 ★/day↗accelerating

Laya exists to give multilingual documents a fast, deterministic decision layer without the latency and parsing drama of autoregressive models.

★ 34.3k Python Language Models · explained Feature
11
deepseek-ai/deepseek-harness
+1468 ★/day↗accelerating

It is an open-source agent harness that treats every component as a plugin, built on an academic framework called Cordis.

★ 261.5k TypeScript Agents · explained Feature
12
Ebony-Vinyl/dsh-our-free-model
+1372 ★/day→steady

A dsh plugin that wires OpenCode's Zen gateway — free, keyless frontier models like Muse Spark 1.3 and MiMo V2.6 — straight into the harness, then spends its engineering budget on the failure modes everyone else skips.

★ 43.1k JavaScript Coding Assistants · explained Feature
13
stablyai/orca
+1318 ★/day↗accelerating

Orca gives every CLI coding agent its own git worktree so you can run Claude, Codex, and Grok in parallel without stashing or branch juggling.

★ 106k TypeScript Coding Assistants · explained Feature
14
Graphify-Labs/graphify
+986 ★/day↗accelerating

Graphify builds a queryable knowledge graph from your whole project so AI assistants can answer architecture questions without grepping raw files.

★ 135.4k Python Coding Assistants · explained Feature
16
affaan-m/ECC
+799 ★/day↘cooling

ECC exists because raw agent harnesses lack persistent memory, token discipline, and security guardrails, so it layers reusable skills, session hooks, and audit rules on top of them.

★ 288.9k JavaScript Coding Assistants · explained Feature
17
heygen-com/hyperframes
+781 ★/day↘cooling

HyperFrames turns HTML, CSS, and seekable animations into deterministic MP4s, built for agents and developers who'd rather skip React and proprietary timelines.

★ 72.2k TypeScript Coding Assistants · explained Feature
18
pbakaus/impeccable
+729 ★/day↗accelerating

Impeccable exists to stop AI coding assistants from defaulting to the same handful of SaaS-template tells—Inter everywhere, nested cards, purple gradients—by giving them a curated design vocabulary and 27 deterministic anti-pattern rules.

★ 87.8k Rust Coding Assistants · explained Feature
19
anomalyco/opencode
+702 ★/day↗accelerating

It gives your terminal an AI pair programmer that can either build freely or explore cautiously without touching your files.

★ 221.6k TypeScript Coding Assistants · explained Feature
20
ifixai-ai/iFixAi
+693 ★/day↘cooling

iFixAi runs up to 32 inspections against any LLM or agent and returns a letter-grade scorecard in minutes, using a separate provider as judge so the model isn't grading its own homework.

★ 34.4k Python LLMOps · Eval · explained Feature
loading more…

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.