An open-source toolkit for testing and evaluating prompts, LLMs, and vector databases, letting developers run experiments from code, notebooks, or a local playground.hegelai
OpenLIT is an open-source, OpenTelemetry-native observability platform for AI and coding agents. It provides tracing, evaluations, guardrails, prompt management, and cost and GPU monitoring, and can be self-hosted.openlit
Lightweight Python tool that cleans AI refusal messages from Codex CLI session files and repairs context, helping developers maintain usable conversation history for coding sessions.ryfinez
An open-source LLM inference engine tuned for AMD consumer hardware, with speculative decoding, long-context tricks and OpenAI/Anthropic-compatible APIs that coding agents can connect to.luce-org
Project Maya is a local LLM inference engine that runs large GLM, DeepSeek, and MiniMax models on your own GPUs, with quantized weights, a web dashboard, and OpenAI/Anthropic-compatible APIs.mw00
A cross-platform desktop panel for installing, configuring and operating multiple AI agent engines with chat, models and memory in one place.qingchencloud
A blazing-fast open-source LLM inference engine tuned for agentic workloads, pairing TensorRT-class speed with vLLM-like ease for production deployments at scale.lightseekorg
Piston is a high-performance, general-purpose code execution engine that runs untrusted code safely inside isolated sandboxes and exposes a simple HTTP API with dozens of language runtimes.engineer-man
LayerFS is an agent filesystem substrate: every agent gets a writable ephemeral workspace over shared committed history, with commits, branches, rewind, merge and content deduplication.Ephemeral-AI-Lab
Tirith is a terminal security gate for developers and AI agents. It blocks homograph URLs, pipe-to-shell, ANSI injection, data exfiltration and malicious skills before execution, adding shell hooks, an MCP gateway and sandboxed runs.sheeki03
A security scanner for AI Agent Skills that combines pattern rules, AST dataflow analysis, an LLM judge and a CEL decision layer to detect prompt injection, data exfiltration and malicious code.cisco-ai-defense
AgentField is a production control plane for AI backends that turns agent functions into REST services, adding orchestration, queues, retries, identity, audit trails, and observability.Agent-Field
Dependency-free, embeddable C inference engine for ultra-large mixture-of-experts models, streaming activated Kimi K3 weights from NVMe with paged caching to run beyond available RAM.sqliteai
A middleware proxy that turns the Google AI Studio web UI into an OpenAI-compatible API through Playwright/Camoufox automation, with function calling, auth rotation, streaming and a built-in web UI.CJackHwang
A collection of free tools that proxy the Claude Code CLI to free NVIDIA-hosted models such as Kimi, GLM, Qwen and DeepSeek, distributed as ready-to-run executables.cixingguangming55555
An HTTP credential proxy and vault for AI agents that brokers secrets on outbound requests, so agents can call APIs without ever holding real credentials.infisical
A platform for building tested agent skills and governing their lifecycle through a team marketplace, with approvals, versioned releases, rollback, quarantine, and multi-platform distribution.FrancyJGLisboa
OpenWolf gives coding agents like Claude Code, Codex, and OpenCode a shared local project memory, code index, task handoff, and token-usage accounting, all without extra model calls.cytostack
Muxy is a lightweight, memory-efficient terminal multiplexer with split panes, background and remote sessions, git worktrees, and integrations for AI coding agents.muxy-app
A proxy gateway that exposes Kiro (Amazon Q Developer / CodeWhisperer) accounts as OpenAI- and Anthropic-compatible API endpoints, letting coding clients use Claude and other models through a subscription you already have.jwadow
ToolHive is an open-source MCP platform that runs MCP servers in isolated containers and adds a gateway, registry, Kubernetes operator, policy enforcement and observability.stacklok
An offline desktop app for browsing and searching the session history of Claude Code, Codex, Cursor and other AI coding assistants, with transcript review, cost insight and an optional self-hosted WebUI.jhlee0409
A desktop tool for browsing and searching past sessions from Claude Code, Codex, Cursor, and other coding agents, with statistics and export for reviewing your work.jhlee0409
According to its description, a relay that aggregates hundreds of free AI model quotas and plugs them into local projects; the README is unavailable, so its exact features remain unverified.romgX
A native workspace for terminals and AI agents that manages repositories, worktrees, tabs and agent sessions, with change review, SSH hosts and multi-machine support.sheprd-labs
Sheprd is a native desktop workspace for terminals and AI coding agents, organizing repositories, worktrees, sessions, reviews, and agent activity in one app.penso
An experimental Windows Terminal fork with native ACP agent integration, offering a docked agent pane, session management, error detection and support for multiple coding-agent CLIs.microsoft
A local-first token usage and cost tracker for dozens of AI coding tools, with native desktop apps, widgets, quota monitoring and optional cross-device sync.xiufengsun
Cate is an infinite-canvas IDE for parallel coding agents. Terminals, editors, browsers, and agent panels float in one workspace, with worktrees, session resume, and shared remote access.0-AI-UG
A local API bridge that lets coding clients like Pi and OpenCode use your Claude and Antigravity subscriptions, with a usage dashboard and an optional Mac app for provider management.rynfar
The official Windows companion for OpenClaw, combining a tray chat client, gateway connection and a node mode that safely exposes shell, browser, camera and screen capabilities to agents.openclaw
Agent Safehouse is a macOS sandboxing layer that confines local LLM coding agents with deny-first policies, so they can only read, write and reach the files and integrations they actually need.eugene1g
A local-first desktop workbench for agent builders: author and version prompts and tools, trace every model call and tool run, replay failures, and evaluate agent performance in one place.deer-flow
claude-code.nvim integrates the Claude Code AI assistant into Neovim, letting you toggle it in a terminal window and automatically reload files it edits.greggh
Open Island is a native macOS companion that puts live coding-agent sessions in your menu bar or notch, showing status, handling permission approvals and jumping back to the right terminal.octane0411
A browser and phone client for herdr that turns coding-agent sessions into chat cards, with live terminal access, remote SSH machines, and push alerts.devswha
Hermes Desktop is a native macOS companion for Hermes Agent that connects locally or over SSH to manage sessions, workflows, Kanban, skills, cron jobs, usage, and a real terminal.dodo-reach
Ovis is a browser and phone client for herdr that lets you follow and answer Claude Code, Codex and other coding-agent sessions remotely, with live terminals and push alerts.devswha
A daemon that bridges messaging platforms such as Feishu with AI coding CLIs, spawning an isolated session per topic, streaming output back and supporting multi-bot collaboration and web terminals.deepcoldy
DeepSWE is a benchmark for measuring frontier coding agents on original, long-horizon software engineering tasks. It includes 113 tasks across five languages, isolated environments, program-based verifiers and tooling for reproducible evaluations.datacurve-ai
TensorFold is a local LLM inference engine for Apple Silicon and NVIDIA GPUs. It loads checkpoints and serves models through an OpenAI- and Anthropic-compatible API without Python.ashhart
Self-hosted gateway exposing Grok chat, image and video capabilities through OpenAI-compatible APIs with multi-account pooling and rotation for stable access.jiujiu532
An optimized quantization and inference library for running local LLMs on consumer GPUs, with EXL3 quantization, tensor and expert parallelism, CPU offloading, and OpenAI-compatible serving.turboderp-org
dmux is a terminal multiplexer for coding agents that gives each task its own git worktree and tmux pane, so multiple AI agents can work in parallel and merge results safely.standardagents
A management console for the Null ecosystem that installs, configures, supervises, and monitors AI agents and orchestration components from one web UI.nullclaw
voxl is a low-latency Rust desktop streaming server that sends headless VM or sandbox screens to a browser over WebSocket, with screenshot and screen-wait APIs built for computer-use agents.tembo
Zilla is a streaming-native, multi-protocol gateway that routes LLM requests across providers, governs MCP tool access and bridges Kafka, HTTP, gRPC and other protocols in one declarative runtime.aklivity
An open-source benchmark for legal work that evaluates LLM agents on realistic tasks with datasets, execution harness, scoring rubrics, and reports.harveyai
Wake is a native desktop app that collects coding-agent sessions from many tools into one searchable, resumable timeline, with MCP and CLI access.iamcorey
An integration exposing DeepSeek V4 models inside the Copilot Chat model picker while preserving agent mode and tool calling, with API keys stored in the system keychain.vizards
A three-tier framework from NVIDIA for auditing AI agent skills: static quality and security checks, semantic overlap detection, and live agent evaluations that measure how a skill changes agent behavior.nvidia
A GPU-rendered terminal workbench written in Rust with persistent sessions, remote workspaces, and agent-aware features that let one coding agent launch another, hand off tasks, and read results.l0ng-ai
Open-source observability platform aggregating traces, logs, and metrics with grouped alerting, using AI agents to help diagnose issues and support self-healing workflows.superloglabs
A native macOS terminal workspace combining tabs, splits, a browser, and a file tree, with background AI agents and unified state management for collaborative command-line work.egoist
A self-hosted compression gateway for coding agents that shrinks prompts and tool history while keeping originals recoverable on demand.Paritok-official
A desktop and server workspace for AI coding agents, running Claude, Codex, Cursor, and others in parallel Git worktrees with shared session, terminal, diff-review, and repository integrations.coollabsio
A desktop IDE for AI coding agents that manages multiple Claude Code, Codex, and Gemini CLI sessions in parallel, with cost tracking and session resume.elirantutia
TraceRoot is an open-source observability and continuous-improvement platform for AI agents that traces model and tool calls, detects production failures, clusters signals and verifies fixes with evals.traceroot-ai
A library for evaluating and improving models and agents through environments, providing environment tooling, scalable evaluation, verifiers, and reinforcement-learning training support.nvidia-nemo
GoModel is a self-hosted AI gateway written in Go, unifying OpenAI- and Anthropic-compatible APIs with multi-provider routing, caching, budgets, observability, guardrails, and usage analytics.ENTERPILOT
Open installer scripts that connect Codex CLI and Claude Code to the NeuroAPI multi-model service, handling API keys, profiles, and client configuration on Windows and macOS.neurogen-dev
AI Runner is a headless offline inference service for art, speech, LLM chat, and automated workflows, exposing model workers and HTTP/WebSocket APIs to other applications.Capsize-Games