apache2-licensed
The LLM Red Teaming Framework
Brought to you by the team behind DeepEval, DeepTeam enables teams to red team any AI application — surfacing vulnerabilities, jailbreaks, and agentic risks before attackers do.
tiny Jev-like family of decision models built on top of Qwen3.5 you can train and run on your own.
One call. not ten reads. code intelligence for your AI coding agent.
Gortex indexes your repository into an in-memory knowledge graph and serves it to your coding agent over MCP, HTTP, and a web UI. One graph query replaces a stack of file reads — up to 50× fewer tokens per response. 257 languages, a 21-tool MCP surface, live editor-buffer overlays, speculative edits, dataflow, clone detection, graph-grounded PR review, and a 17-server LSP bridge. Built for 19 AI coding agents on macOS, Linux & Windows — one binary, zero services.
Knot is a platform for managing development environments, testing systems, automation workflows, and platform services across containers and virtual machines. Create, manage, and share everything from disposable sandboxes to long-running workloads through a single interface, whether running locally or across a cluster.
Rust Glancer is an experimental LSP implementation that tries to use a different architecture compared to rust-analyzer in order to lower RAM usage and make editor restarts faster.
Related contents:
Google's open agentic orchestrator. Declare an agentic task. AX runs it at scale.
AX sandboxes your task, wires up its workspace, fences its network, and helps you run billions of them per cluster. Either use a single task per agent, or compose as many as your agent needs.
Multilingual, non-autoregressive System 1 decision engine. Typed decisions over 100+ languages in a single forward pass — 33 ms — trained with reinforcement learning against strictly proper scoring rules (RLCD), with a router that picks the right checkpoint per request.
Fault-tolerant, highly scalable GPU orchestration, and a machine learning framework designed for training models with billions to trillions of parameters.
OpenDepot is a self-hosted, Kubernetes-native registry where Depots continuously reconcile OpenTofu and Terraform modules and providers. Define distribution in manifests and let OpenDepot keep your registry in sync.
Related contents:
a fast language that blocks AI mistakes via proof
C speed · CUDA parallelism · Lean proofs · Python syntax
In the post-AGI economy, humans will eventually stop writing and reading code, but we still need an ambiguity-free way to tell the AIs building the world around us what we want done.
With laws, our intents can be much more precise than natural language. With proofs, we can verify that the AI implemented our prompts correctly. And a fast compiler runs it at speed.a fast language that blocks AI mistakes via proof
C speed · CUDA parallelism · Lean proofs · Python syntax
In the post-AGI economy, humans will eventually stop writing and reading code, but we still need an ambiguity-free way to tell the AIs building the world around us what we want done.
With laws, our intents can be much more precise than natural language. With proofs, we can verify that the AI implemented our prompts correctly. And a fast compiler runs it at speed.
Revisiting Diffusion Transformers for Monocular Depth Estimation.
Marigold V2 is a family of models and a cost-effective fine-tuning protocol that repurposes a pretrained diffusion transformer into single-step dense predictors: depth, see-through depth, surface normals, albedo, and other dense modalities. Fine-tuning takes less than a week on a single consumer GPU, within reach of individual practitioners and small labs, and the results are state of the art, faithfully reproducing sharp edges, fur, and hair-thin details. The same models also unlock applications such as metric depth completion.
A self-contained coding agent.
Ghost in your shell. Ante is a self-contained agent harness with a highly optimized core. It works like Claude Code or Codex, with none of their dependencies or model constraints.
build, expect, train, x-ray, merge, bisect, ship, one CLI.
Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done. Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.
Effective Linux Compliance
Pavois audits the configuration your services actually run, sshd -T, sysctl, systemctl show, not just the files on disk. It catches the Includes and drop-ins that file-based scanners miss, grades the result A:E, and remediates it as code.
A local-first inbox for long-running AI agents, built with DeepAgents and LangGraph.
Pizza Bot is an inbox for long-running AI work. Start or schedule a task, return to your day, and let completed work collect in Unread while runs waiting for your decision collect in Action. Agents keep working when you navigate away or disconnect; the api-server process must remain running.
task orchestration for coding agents. One goal. An ordered plan of agent tasks. Every runner, model and effort visible before you spend a token.
The planner reads your repo, asks about whatever you left vague, and hands back tasks you can rewrite — prompts, models, modes, dependencies, all of it. Then it runs one real coding-agent session per task, and completes a task only when that task's marker shows up in the output.
Multi-agent task orchestration for coding agents. Turn one goal into an ordered plan of tasks — each with its own runner, model and mode — then execute and verify the results.
a continuous-improvement harness for long-running agent tasks. Turns Claude Code, Codex, or Hermes into a multi-agent mission orchestrator via MCP/ACP.
Zenith is an agent harness for work that may run for days or weeks, where the dominant failure mode is premature completion rather than inability to make progress. It runs a coding agent (Claude Code, Codex, or Hermes) as a multi-agent orchestrator over MCP/ACP: one orchestrator session reads task state each turn and decides whether to spawn workers and testers, register reusable skills, replan, or stop.
Charts built for Chat
AI agents make an unauditable mess of dashboards. Give them dbt Charts, a simple declarative YAML language, and they build better dashboards you can govern, using fewer tokens.
Related contents:
LAN visibility and alerting tool for home networks and small offices. Self-hosted network visibility for discovering, organizing, and monitoring devices on your LAN.
LanGuard finds devices, tracks online and offline state, scans common ports, keeps network history, and can send Discord, Telegram, or automation webhook alerts when new devices appear.
a Kubernetes TUI that tells you why it’s broken. The cluster tells you it’s broken. sofka tells you why.
A Kubernetes terminal interface inspired by k9s. Browse built-in and custom resources, run Flux CD actions, and inspect the evidence behind resource failures.
A de-Googled Android build for the Pixel 10a (stallion). Push-to-talk assistant, your own backend. LLM assistant-first phone firmware for the Google Pixel 10a.
A de-Googled phone OS built on GrapheneOS for Google Pixel 10a. It is a minimalist, assistant-first OS designed to work with your own backend. It has a small fixed set of local apps including the phone, messages, camera, gallery, offline maps, flashlight and settings, and nothing else. No app store, no browser, no feed, no Google account, and no way to add anything.
Related contents:
High-Quality Voice Cloning TTS for 600+ Languages.
OmniVoice is a state-of-the-art massively multilingual zero-shot text-to-speech (TTS) model supporting over 600 languages. Built on a novel diffusion language model-style architecture, it generates high-quality speech with superior inference speed, supporting voice cloning and voice design.
Related contents:
Durable, searchable memory of your past agent sessions.
Durable memory for your AI coding agents. funes indexes your past sessions across Claude Code, Codex, pi, and Hermes and lets any agent recall the past decisions, rationale, and findings. Your memory is a dataset you can publish to the Hugging Face Hub — then any machine, teammate, or agent can recall from it.
Related contents:
An IDE designed for agentic AI research. AgentsDock currently supports Claude Code, Codex, and Cursor in one desktop and mobile workspace.
157 cinematic shot recipes. One prompt to a finished promo. AI Video Skill for Claude Code & Codex. Cinematic Product Videos with Remotion.
Every motion you see behind this page is a tuned, ready-to-use shot recipe card — your agent picks the shots, animates real product screenshots with Remotion, syncs the beats, and delivers the film.
Portable Toolkit for Building Secure Software.
Mantis is a set of skills along with an ADK reference harness for building secure software in the new AI era of software development.
A modular, stack-agnostic toolkit of security review skills for AI coding agents to autonomously find, reproduce, and patch vulnerabilities.
A Compose control plane that does not own your containers.
Windlass gives you projects, Git deployments, HTTPS domains, logs, terminals and backups on your own server. It shells out to the real docker compose rather than replacing it, so removing the panel leaves everything running.
Review the package artifact before it ships. pre-publish package security. Review the artifact before you publish it: diffs the built npm/PyPI package against the last published version and pins risk signals to the changed lines.
Between your last code review and the public registry sit build scripts, bundler output, and CI credentials. Drydock diffs the exact artifact against the last published version and pins every supply-chain finding to a changed line. Workflow Gate enforces the decision on a configured protected job; Stage Watchtower records an advisory npm review.
Your Logitech mousekeyboardcamera, finally local.
A local-first alternative to Logitech Options+, written in Rust.Remap buttons, drive DPI and SmartShift over HID++.No account, no telemetry.
Related contents:
photos without a server.
Smriti indexes any drive, recognises faces on-device, and surfaces memories — without leaving your machine. No account, no cloud, no server humming in a closet.
Related contents:
Cursor for circuit boards.
Copperhead is an open-source AI engineering platform that helps hardware teams design, verify and ship circuit boards.
An agent-first decompiler designed to be refined by other agents. Kuna is written in Rust and was originally ported from Ghidra.
Kuna is a self-refining decompiler built to be used by agents. It is based on the Ghidra decompiler, written in Rust, and integrates modern features from other open-source decompilers and research. Kuna is an experiment to see how far agents can take self-refining tools.
Safety for Agents. Agent Detection & Response for AI coding assistants .
Lightweight Agent Detection & Response (ADR) layer for AI agents — guards commands, files, and web requests. Part of Gen Agent Trust Hub.
Sage is a lightweight security layer that protects AI agents from executing dangerous actions. It intercepts tool calls — shell commands, URL fetches, file writes — and checks them against multiple threat detection layers before they run.
Offensive security for agentic infrastructure. The red team framework for AI agent infrastructure.
Drop one static collector on a compromised host. One active scan captures configs and raw credentials, discovers MCP, A2A and AI services, reuses compatible secrets, proves concrete access, and checkpoints the evidence before the foothold disappears. Ingest later for full graph analysis.
Offensive security framework for AI agent infrastructure - recon, credential looting, model exfiltration, poisoning, and attack-path analysis across MCP, A2A, gateways, and AI services. BloodHound for the agentic stack.
Keep private data, internal infrastructure and secrets out of cloud coding agents without breaking your workflow.
Cover is a bidirectional privacy proxy for AI coding agents. It replaces matched sensitive values locally with realistic, deterministic stand-ins before a request leaves your machine, then translates matching fakes in normal and streaming responses back to the originals. The model gets coherent context; your agent and tools keep working with the real environment.
Give every agent its own browser.
The headless browser for AI agents and web scraping. Obscura is a headless browser engine written in Rust, built for web scraping and AI agent automation. It runs real JavaScript via V8, supports the Chrome DevTools Protocol, and acts as a drop-in replacement for headless Chrome with Puppeteer and Playwright.
Execute chDB queries in Postgres.
This library provides PostgreSQL extensions for executing chDB queries in Postgres, and for copying data from external sources into a PostgreSQL table.
Related contents:
Router that virtually distributes inference across connected devices in the home.
NVIDIA Personal AI Router (PAIR) is a local inference router for a group of compatible computers on the same network. It discovers participating nodes, manages supported inference engines, and presents Ollama-compatible and OpenAI-compatible proxy endpoints to applications and agents. Independent requests can be routed to eligible nodes according to engine availability, model availability, and current workload.
PAIR is useful for concurrent local workloads such as multi-agent applications. Prompts and responses are intended to remain on the local network when every configured client, model source, engine, and node is local.
27 Security Intelligence Tools for Claude.
Production-grade MCP server giving Claude 27 security intelligence tools across 21 APIs — CVE lookup, EPSS scoring, CISA KEV, MITRE ATT&CK, Shodan, VirusTotal, and more.
Semantic Versioning for Nix Flakes.
Thaw works by operating directly on the ref specified for your flake's inputs. If a ref exists on the input and is a valid SemVer version, thaw will attempt to upgrade it to the latest version. Both url querey param and object configuration syntax are supported.
By default, non-major upgrades are allowed, but this can be changed with the --major flag. No additional tools or components are required.
A hive mind communication platform.
A workspace where humans and agents build together, on a relay you own.
Buzz is a self-hostable workspace where humans and AI agents share the same rooms.
A Buzz community is the workspace a user reaches by URL. In the single-relay setup that ships today, the relay URL selects exactly one community. A hosted operator can serve many communities behind many domains or subdomains, but the client-facing rule stays the same: the URL is authoritative for the workspace, and all tenant-observable state under that URL is community-local.
Safe schema management for MySQL and MariaDB.
Skeema is a declarative SQL-based schema management utility. Simply track your desired schema state in a repo of CREATE statements, and the tool figures out how to apply any changes to your tables and routines. You won't need to code migrations ever again.
infrastructure as code, in Lean 4.
Terraform-style declarative infrastructure, written in the Lean programming language and proof assistant instead of a bespoke DSL. Target and observed cloud state are dependently-typed Lean values, so an unrealisable target is a compile error rather than a failed apply.
A dangling reference, a region a cloud is not in, an instance size that does not exist, a plaintext password in a committed file — none of these get as far as a plan.
Related contents:
Rust LSP that doesn't eat memory for breakfast.
Rust Glancer is a memory efficient LSP for Rust. It uses a different approach compared to rust-analyzer: instead of storing everything in memory and recomputing dynamically, it uses frozen workspaces that can be offloaded to the filesystem. As a result, the LSP gets some performance penalty, but is able to achieve extreme memory efficiency and instant restarts.
Own your logs. Search them fast.
Rootprint is open-source, self-hosted log management with OpenTelemetry-native ingestion, full-text search, and traces next to the logs that emitted them. You pay for retention at your object storage's $/GB.
A fast, open-source Postgres terminal written in Rust. 🦀🐘 Connect and monitor all your databases from one place — with pgbot built in for deeper diagnostics.
htop for all your Postgres databases — an interactive terminal UI that monitors every database you care about in one place, powered by pgbot's read-only diagnostics.
A Kubernetes operator of virtual BMCs that provide Redfish and IPMI services for KubeVirt virtual machines.
Bare-metal provisioning tools like Metal3, Tinkerbell, and Seeder rely on IPMI and Redfish to manage server power state, boot devices, and firmware. KubeVirt virtual machines have none of these interfaces out of the box, making them invisible to existing infrastructure automation.
KubeVirtBMC fills that gap. It is a Kubernetes operator that creates a virtual BMC (Baseboard Management Controller) for each KubeVirt VM, exposing standard IPMI and Redfish endpoints that map directly to Kubernetes-native operations. The result: your VMs become first-class citizens in any bare-metal provisioning pipeline.
Related contents:
Run your agent on local models. Free, private, and offline.
Open source inference server that runs the best local models for your hardware, plugged into the agent you already use. Works with Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, Oh My Pi, and Cline.
High-Performance In-Browser LLM Inference Engine.
WebLLM is a high-performance in-browser LLM inference engine that brings language model inference directly onto web browsers with hardware acceleration. Everything runs inside the browser with no server support and is accelerated with WebGPU.
Efficient and versatile WebAssembly interpreter for embedded systems.
Wasmi is an efficient and lightweight WebAssembly interpreter with a focus on constrained and embedded systems.
Related contents:
A fast, simple, and embedded key-value store for Rust.
TurboKV is an async embedded key-value database with atomic batches, ordered range scans, configurable durability, compression, and background compaction.
The open source AI gateway.
An open source model gateway that provides one control plane across closed, open-source, local, and custom models. One API key for every model: hosted providers, your own keys, your own GPUs. Routing, access control, and attribution for everything your agents call.
AWS Bloodhound OpenGraph Connector.
awshound collects AWS IAM/authorization data and builds a BloodHound OpenGraph.
Related contents:
Kafka Load Testing & Chaos Engineering. Break Kafka before production does.
Khaos generates realistic Kafka traffic and reproduces the failure and performance conditions that only show up under real load: consumer lag, hot partitions, rebalances, and broker failures.
AI-powered virtual executive team — a single coherent executive persona backed by 8 specialist Claude agents (FastAPI + Next.js).
rootless container sandbox that starts in 3.5 ms.
A fast, rootless sandbox and virtual resource runtime for any workload, including untrusted and AI-generated code. One 1.52 MB static binary, no daemon, 0 RAM at rest.
One binary that manages resources, of which isolation is the first. That is why there is no single row for kern in a comparison table: it is a container runtime, a sandbox, a resource slicer and a stack runner at once, in 1.52 MB with no daemon.
Durable streams on Amazon S3.
PicoMQ is durable, real-time streams over HTTP, built on S3-compatible object storage.
PicoMQ is a durable stream server. Clients create named streams, append records, and read them back over HTTP. Records are stored on S3-compatible object storage and cluster coordination goes through a SQL database. A node is a single binary with no local state worth backing up.
Secure code execution for AI applications.
run executes untrusted JavaScript and type-stripped TypeScript in a hardened QuickJS sandbox. Guest code can access only the host functions you provide.
Related contents:
An LLM tool-use loop (bash / read / write / edit / subagent), an interactive bubbletea session, and provider-routable models. One binary, no runtime, config you can read.
A fast coding-agent harness in Go. Tool-use loop, bubbletea TUI, provider-routable models with live catalog discovery, MCP support, background subagents. One binary, no runtime, built for open-source models.
Local-First AI Agent Workspace.
A local-first Agent workspace built for real work. Maka inspects projects, runs tools under a sandbox boundary, and records model messages and tool calls as recoverable execution facts — on your machine, through one Runtime Host.