Open-Source AI Tools
Open-Source AI tools on PulseGate. 6,876 AI tools are open-source, most recently sentinel-ai-auditor.
sentinel-ai-auditor is a deterministic, offline-first security auditor for AI agent skills and instruction bundles.
reprosieve reduces failed agent traces into redacted offline predicate reproductions.
mission-directives provides deterministic prompt routing, project context, and verifiable agent workflows for AI agents.
multigrid-sports provides multi-agent gridworld environments for RL research.
TimeOmni-1-7B is a 7B parameter language model with tool-calling capabilities based on Qwen.
agent-voice provides local, free voice artifacts and TTS for AI agents.
attestroot is a local-first, tamper-evident evidence layer for AI agent activity using hash-chained capture and Merkle commitments.
CrewScore provides an offline structural production-readiness scorecard for AI agent system prompts.
Lee-AI is a live AI cursor that talks, points, frames products, and operates websites for visitors.
qwen2.5-1.5b-instruct-q4_k_m.gguf is a 4-bit quantized GGUF version of the Qwen2.5 1.5B Instruct model.
GLM-4.7-Flash-GGUF is a GGUF-quantized version of the GLM-4.7-Flash language model for local inference.
musicgen-large is a large text-to-music generation model developed by Meta.
hrnet_w18.ms_aug_in1k is a High-Resolution Network (HRNet) image classification model trained on ImageNet-1K.
xlm-roberta-large-xnli is an XLM-RoBERTa model fine-tuned for zero-shot text classification using the XNLI dataset.
Qwen-Marketing is a Qwen model fine-tuned for marketing-related tasks and content generation.
swin-base-patch4-window12-384-in22k is a Swin Transformer model pretrained on ImageNet-22k for image classification.
bert-base-chinese-ws is a BERT model fine-tuned for Chinese word segmentation.
deepseek-r1-distill-qwen-32b-awq is an AWQ-quantized distilled version of a DeepSeek reasoning model based on Qwen.
phobert-base-v2 is the second version of a RoBERTa-based model pretrained on Vietnamese text.
tiny-random-BertModel is a minimal random-weight BERT model used for testing and debugging.
inception_resnet_v2.tf_in1k is an image classification model combining Inception and ResNet architectures, trained on ImageNet.
opus-mt-mul-en is a multilingual-to-English machine translation model from the OPUS-MT project.
bert-base-turkish-cased-ner is a BERT model fine-tuned for named entity recognition in Turkish.
Qwen3-ASR-0.6B-8bit is an 8-bit quantized automatic speech recognition model optimized for MLX.
Multimodal model from Moonshot AI that processes both images and text.
degardis compiles validated Agent Skill sources into portable, installable bundles.
DiffusionGemma is an open-source text-to-image diffusion model based on Gemma, optimized for NVIDIA hardware.
critic-orchestrator is an MCP server that spawns parallel adversarial AI reviewers to verify claims and reduce confabulation in AI coding agents.
llm-perplexity-agent provides Perplexity's Agent API as a model plugin for Simon Willison's llm CLI tool.
csoai-defoneos-mcp is an MCP server providing sovereign UK defence-AI certification surface with multi-agent governance and audit capabilities.
notionmemory is a Python library providing Notion as a second-brain hub with memory, calendar, templates, and skills for coding agents.
skilldistill turns successful AI agent sessions into reusable agent skills by distilling Claude Code transcripts into SKILL.md playbooks.
Compartment provides high-security, fully offline, encrypted vector memory for AI agents using MCP.
Tokenbill is a Python package for profiling LLM agent traces to identify token waste, prompt cache issues, and cost drivers.
Dunnit is a Python package for tamper-evident verification of AI agent work using definition-of-done as code.
idios-mcp-server provides private escrow services for AI agent work on Beam.
videoprompt is a CLI tool to plan, generate, edit, and validate multi-shot videos using Grok and xAI.
archery-mcp-server allows AI agents to directly operate SQL tickets, queries, and audits via the Archery platform.
syncanything provides an agent-native context index that works across multiple AI products.
agent-trajectory normalizes native agent transcripts into a shared trajectory format.
agentmandate provides compound-path and cross-release analysis of AI agent permissions and policies.
Fara1.5-27B is a Microsoft fine-tuned 27B parameter model based on Qwen for computer use.
Agents-A1-4B is a 4B parameter language model specialized as a deep research assistant.
Mamba2-primed-HQwen3-8B-Instruct is a hybrid Mamba-Transformer model primed with Qwen3 for instruction following.
Qwen3.6-35B-A3B-GGUF is a GGUF quantized version of the Qwen3.6-35B model for efficient local inference.
Ondaru provides agent-ordered 3D product mockups served as an A2MCP service on OKX.AI.
meeting-transcribe converts meeting recordings into speaker-labelled transcripts and evidence-grounded summaries.
flightdeck-verify is a standalone offline verifier for FlightDeck evidence packs generated by AI agent fleets.
Evarness is a Python package for proving AI agent harnesses with traces, invariants, and offline verification.
Stepproof is a Python package providing step-level verification and tamper-evident audit trails for AI agents.