Official add-on registry for Crow — installable services, MCP servers, skills, and panels.
| Add-on | Type | Description |
|---|---|---|
| Google Workspace | MCP Server | Drive, Docs, Sheets, Slides, Gmail, and Calendar with your own Google account |
| Obsidian Vault | MCP Server | Connect your Obsidian vault for reading, searching, and syncing notes |
| Home Assistant | MCP Server | Control your smart home — lights, temperature, switches |
| Ollama | Docker Bundle | Run local AI models for embeddings and summarization |
| Nextcloud | Docker Bundle | Self-hosted file sync and collaboration via WebDAV |
| Immich Photos | Bundle | Search photos, browse albums, manage your Immich library |
| Podcast Publisher | Skill | Publish podcast episodes with iTunes-compatible RSS feeds |
Bundles below run local LLMs via llama.cpp (Vulkan AMDVLK) or vLLM (ROCm/CUDA). Most bundles require a pre-downloaded GGUF/HF model; see each bundle's docker-compose.yml comments for the exact curl command. All reference shared env files at ~/.crow/env/rocm.env or ~/.crow/env/cuda.env (operator-maintained; see the crow main repo CLAUDE.md § "Adding a new bundle add-on").
| Add-on | Hardware | Description |
|---|---|---|
| llamacpp-vulkan-qwen36-35b-a3b | AMD Strix Halo gfx1151 | Qwen3.6-35B-A3B (35B/3B active MoE) UD-Q6_K + mmproj-F16 vision encoder; agentic coding + VLM. Apr 2026. |
| llamacpp-vulkan-qwen3-coder | AMD Strix Halo gfx1151 | Qwen3-Coder-30B-A3B Q8_0 MoE, on-demand code specialist. |
| llamacpp-vulkan-glm-45-air | AMD Strix Halo gfx1151 | GLM-4.5-Air 106B/12B MoE Q5_K_M, deep-reasoning specialist. |
| llamacpp-qwen72b | AMD Strix Halo gfx1151 | Qwen 72B via llama.cpp. |
| vllm-rocm-qwen3 | AMD Strix Halo gfx1151 | Qwen3-4B via vLLM-ROCm, fast dispatch model. |
| vllm-rocm-qwen3-32b | AMD Strix Halo gfx1151 | Qwen3-32B via vLLM-ROCm, daily-driver mid-tier. |
| vllm-rocm-qwen35-4b | AMD Strix Halo gfx1151 | Qwen3.5-4B via vLLM-ROCm. |
| vllm-rocm-qwen35-27b | AMD Strix Halo gfx1151 | Qwen3.5-27B via vLLM-ROCm. |
| vllm-rocm-kimi | AMD Strix Halo gfx1151 | Kimi model via vLLM-ROCm. |
| vllm-cuda-embed | NVIDIA CUDA | Qwen3-Embedding-0.6B BF16 for semantic memory/search. |
| vllm-cuda-rerank | NVIDIA CUDA | Qwen3-Reranker-0.6B BF16 cross-encoder for top-K reranking. |
| vllm-cuda-vision | NVIDIA CUDA | Qwen3-VL-4B-Instruct-FP8 vision-language model. |
From the Crow's Nest Extensions panel, browse and install add-ons with one click.
Or ask your AI:
"Install the Ollama add-on"
- bundle — Docker Compose services managed by the
crowCLI - mcp-server — MCP servers installed via
npxoruvx, registered in.mcp.json - skill — Markdown behavioral prompts copied to
~/.crow/skills/ - panel — Crow's Nest dashboard panels placed in
~/.crow/panels/
registry.json is the master manifest fetched by the Crow's Nest Extensions panel. Each add-on directory contains a crow-addon.yml with metadata, requirements, and configuration.
See the Add-on Registry docs for the full schema.
Anyone can create a community add-on store by following the community store template.
See the submission template for submitting new add-ons.
MIT