Jumper

An open-source set of reinforcement-learning training environments for Pollen Robotics' Microduck bipedal robot, built on mjlab with MuJoCo Warp and PPO. It trains walking, standing, recovery, sitting, ground-picking, kicking, rolling, and roller-skating policies, with flat, rough, slope, and backlash variants where applicable. The environments encode a sim-to-real pipeline using BAM actuator physics for Dynamixel XL330 servos, domain randomization for battery voltage, voltage sag, command delay, and friction, and simulated gear backlash with output-side encoder feedback. Policies use a shared 61-dimensional observation contract for runtime hot-swapping, are trained at 50 Hz, exported to ONNX with the observation normalizer embedded, and deployed by the Microduck runtime. Training runs locally through a CUDA GPU or can be submitted to Hugging Face Jobs; the project is licensed under Apache 2.0, while its 3D model files use CC BY-SA-NC.

jcode is an open-source, Rust-based terminal harness for AI coding agents and LLM workflows, designed for interactive development across multiple sessions. It supports resumable sessions, session search, multi-agent swarm and helper-agent workflows, compatible model-provider integrations, MCP, browser automation, and a self-development mode; installation scripts target macOS, Linux, and Windows. For automatic memory, jcode embeds turns and responses as semantic vectors, searches a graph of stored memories using cosine similarity, and feeds relevant results into the conversation. A memory sideagent can verify and expand retrieval, while periodic extraction stores new memories and ambient consolidation reorganizes entries and checks for staleness and conflicts. Explicit memory tools and traditional retrieval over previous sessions are also available. Its terminal UI includes side panels, diff views, inline Mermaid diagrams, information widgets, custom scrolling, and real-time rendering. The project supplies a Mermaid renderer without browser or TypeScript dependencies and publishes benchmarks covering RAM consumption, startup and input-readiness times, and memory scaling across multiple active sessions.

Bot Crossing is a local web application that visualizes coding-agent threads as an astronaut colony. It reads session records from installed agent harnesses on the user's machine, maps repositories to persistent hex zones, and represents sessions as astronauts whose behavior reflects states such as running, waiting, errored, merged, archived, or inactive. Selecting an astronaut or repository opens its associated thread through the owning harness, while the application can also start sessions, reveal folders, mark threads viewed, and archive them. The application uses per-harness adapters to scan local session files and merge them into a common thread model. It currently documents support for Claude Code, Codex, and Cursor, with adapters for other harnesses able to be added under `server/harnesses/`. Astronaut navigation uses a rasterized navigation grid, A* routing, string-pulling, collision handling, and crowd separation. The browser renders the colony with Three.js techniques including instanced GPU-skinned astronauts, shader-based construction progress, procedural terrain and surfaces, and configurable planets, lighting, and quality settings. Bot Crossing runs locally through a Vite development server or a built Node server, binds to loopback by default, and stores the colony layout in `data/colony.json`. It does not upload session data or use an account; it reads harness records and writes only the colony state plus the documented archive field. The repository requires Node 22.13 or newer, supports macOS, Linux, and Windows, and is licensed under MIT. Bundled Kay Lousberg assets are separately covered by CC0, while bundled Material Design Icons use Apache-2.0.

OpenRig is an open-source, self-hosted multi-agent harness for coordinating AI coding agents such as Claude Code and Codex as one managed team. It uses a local daemon, CLI, terminal UI, MCP server, SQLite, tmux, and runtime adapters. Users define agent topologies in YAML with pods, seats, communication edges, and continuity policies, then use commands such as `rig up`, `rig send`, `rig broadcast`, `rig chatroom`, `rig down --snapshot`, and `rig discover`/`rig adopt` to launch, coordinate, inspect, snapshot, restore, or take over existing sessions. Its MCP tools let agents manage their own topology, while the TUI exposes topology, seat, project, feed, terminal, and system views. OpenRig requires Node.js and tmux; herdr, cmux, and Docker are optional for terminal workspaces and service-backed rigs. It is distributed through the `@openrig/cli` npm package and licensed under Apache 2.0.

NanoJev is an open-source 0.6B parallel-decision model and training pipeline based on Qwen3-0.6B. It accepts states, questions, and candidate sets in a single forward pass and returns complete probability distributions without output-token decoding. Shared decision heads support dynamic-choice distributions over supplied candidates, Boolean proposition probabilities, and ordered score levels with expected values. The repository includes data generation, distribution-loss training, evaluation, persistent serving, and browser replay tools. Its game demonstrations combine model judgments with controller code: maze controllers ask local safety questions and track collisions and explored paths, while Snake controllers filter immediate collisions, plan toward visible food, and use the model to break ties between remaining actions. Checkpoints and datasets are distributed through Hugging Face, and the service exposes repeated batched evaluations through a local HTTP endpoint.