Adapts pstack’s rigorous coding workflows for Claude Code, Codex, Copilot, Pi, and other agent harnesses. Task routing, parallel delegation, and verification help agents produce concise, checkable code.
Runs an external, reviewable coding-agent harness that turns intent into repeatable software work: clarifies requirements, builds reviewed plans, executes in tmux-backed sessions, and collects durable verification. Ships Telegram/Discord delivery, a research REPL, and optional desktop-control tools; beta-stage.
Centralizes indexing and management of local AI coding-agent sessions so you can search, view full context, migrate, resume, and restore conversations across agents and devices. Supports extensible local sources, AI summaries, optional Supabase sync, and Skills management.
Routes Hermes Agent coding tasks to models and skills by work type, then runs independent work units in parallel. Its status system distinguishes reported completion from work verified by tests, review, or CI.
Measures how coding agents explore repositories by asking them to return a ranked, line-level list of code regions relevant to an issue under a fixed line budget. Covers 848 issues across 203 repos and 10 languages; evaluates coverage, ranking, and context-efficiency to isolate exploration quality.
Provides a locally runnable, quantized GGUF release of Gemma 4 12B fine-tuned for Python coding with chain-of-thought distilled from Composer 2.5 and supplemented by Fable 5. Multiple quant options for low‑VRAM setups and execution‑verified training traces. Not safety‑aligned; validate before production.
A quantized 27B coder LLM fine-tuned for repository-level code generation, multi-turn tool calling, and agentic workflows — packaged for local GGUF/llama.cpp deployment with MTP speculative decoding and trace-inversion SFT. Optimized for developer tooling; experimental and not fully safety-validated.
Provides a ruleset and skills that make AI coding agents prefer the simplest correct implementation: reuse existing code, prefer stdlib/native features, and only write minimal new code. Cuts generated LOC, tokens, cost and time while preserving validation and safety.
Manages real tmux-backed terminals and AI agents as draggable nodes on an infinite pan/zoom canvas, with a Trello-style kanban view, persistent sessions that survive restarts, mobile companion support, and a browser Server Edition for self-hosting.
A dense ~9B reasoning LLM optimized for agentic coding and tool-calling that emits explicit chain-of-thought (<think>) blocks and well-formed tool calls. Designed to run on a single 80GB GPU (~19GB bf16), uses self-scaffolding RL and exposes an OpenAI-compatible API.
A 35B mixture-of-experts LLM specialized for agentic coding and tool-enabled code generation, fine-tuned with self-scaffolding reinforcement learning. Supports very long contexts, OpenAI-compatible tool calls, and multiple serving runtimes under an MIT license.
Provides an open-source Mixture-of-Experts coding LLM (397B) optimized for agentic, tool-enabled coding workflows with a 262,144-token context window, OpenAI-compatible API, serving recipes (vLLM/SGLang), and published coding-benchmark results.