AlbumentationsX MCP
AlbumentationsX MCP for batch previews, compare preview runs, segmentation masks, and exports.
Open source Open in the app JSON README (API)
About
AlbumentationsX MCP for batch previews, compare preview runs, segmentation masks, and exports.
Details
- Kind
- MCP servers
- Topic
- No topic detected
- Publisher
- dkosarevsky
- Origin
- official
- Category
- ferramentas
- Transport
- local
- Version
- 1.21.1
- Stars
- 6
- Last push
- 2026-08-14T05:24:41Z
- Repository state
- ativo
- Language
- Python
- License
- AGPL-3.0
- Added
- 2026-08-29 03:02:39
- Updated
- 2026-08-29 03:02:39
- Origin id
io.github.dKosarevsky/albu-mcp
README
# AlbumentationsX MCP Model Context Protocol server for [AlbumentationsX](https://github.com/albumentations-team/AlbumentationsX): inspect datasets, preview augmentations, refine them with visual feedback, and export reproducible pipelines. <!-- mcp-name: io.github.dKosarevsky/albu-mcp --> [](https://github.com/dKosarevsky/albu-mcp/actions/workflows/ci.yml) [](https://pypi.org/project/albumentationsx-mcp/) [](pyproject.toml) [](https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.dKosarevsky/albu-mcp) [](https://skills.sh/dKosarevsky/albu-mcp)  Ask an MCP host for several robustness variants, reject an excessive result such as `too_noisy:high`, compare the adjusted batch previews, and export the accepted pipeline. ## Install ### Claude Desktop [Download the latest `albumentationsx-mcp.mcpb`](https://github.com/dKosarevsky/albu-mcp/releases/latest/download/albumentationsx-mcp.mcpb), install it from **Settings -> Extensions -> Advanced settings**, and select separate image and artifact directories. ### Other MCP Hosts Run the published server with bounded local access: ```bash uvx --from albumentationsx-mcp albumentationsx-mcp \ --allowed-root /absolute/path/to/images \ --artifact-root /absolute/path/to/albu-artifacts ``` `run_first_preview` requires the default `full` or `dataset` capability profile. The smaller `review` profile uses the explicit validate/render fallback in the usage guide, or you can restart with `dataset` or `full`; see [configuration](docs/CONFIGURATION.md). Copyable host configurations are in [the install guide](docs/INSTALL.md). The repository also contains a native Codex plugin bundle. `npx skills add dKosarevsky/albu-mcp` installs agent guidance, not the MCP server. ## First Preview After connecting the server, ask your host: ```text Run the host smoke check. If preview_ready is true, call run_first_preview for /absolute/path/to/images with low intensity and at most 8 images. Show me the contact sheet. When I mention a specific result, call trace_preview_variant before adjusting it. ``` `run_host_smoke_check` returns `preview_ready` and a `preview_request_template`. If resource reads are unavailable, call `get_workflow_example` with `example_id="client-smoke"`. Try the [classification robustness use case](docs/use-cases/CLASSIFICATION_ROBUSTNESS.md), or follow the [First 10 Minutes guide](docs/FIRST_10_MINUTES.md). The `validate_preview_request` fallback, batch previews, and how to compare preview runs are in [Usage](docs/USAGE.md). Use `too_noisy:high` or `exposure_too_weak:medium`, then optionally share one redacted loop through [first-preview feedback](docs/FIRST_PREVIEW_FEEDBACK.md). If setup fails, read `albumentationsx://diagnostics/guide` and call `diagnose_environment` for bounded remediation actions. ## Capabilities - Transform discovery, schemas, recipes, and pipeline validation. - Classification, detection, segmentation, OCR, bbox, mask, keypoint, and dataset-quality workflows. - Deterministic previews, contact sheets, annotation overlays, comparison, ranking, and reports. - Interactive MCP Apps review with a text-only fallback for other hosts. - Structured feedback, tuning sessions, and Python, JSON, or YAML export. - Runtime-aware CPU `torch.Tensor` pipeline validation and guarded Python handoff. - MCP `2026-07-28` plus legacy negotiation; stable agent workflow resources, diagnostics, and contract snapshots. The server does not execute arbitrary Python, fetch remote images, overwrite datasets, or train models. Reads are restricted by `--allowed-root`; generated files stay under `--artifact-root`. ## Integrations - [Official Albumentations MCP guide](https://albumentations.ai/docs/integrations/mcp/) - [Official MCP Registry entry](https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.dKosarevsky/albu-mcp) - [skills.sh agent skill](https://skills.sh/dKosarevsky/albu-mcp) - [Upstream documentation PR](https://github.com/albumentations-team/AlbumentationsX/pull/289) ## Documentation - [Install and host configuration](docs/INSTALL.md) - [Runtime settings and capability profiles](docs/CONFIGURATION.md) - [First 10 minutes](docs/FIRST_10_MINUTES.md) - [First-preview feedback](docs/FIRST_PREVIEW_FEEDBACK.md) - [Usage](docs/USAGE.md) and [recipes](docs/RECIPES.md) - [CPU Tensor Compose validation and export](docs/TORCH_CPU_COMPOSE.md) - [MCP Apps review](docs/MCP_APPS_REVIEW.md) and [compatibility policy](docs/COMPATIBILITY.md) - [Documentation index](docs/INDEX.md) - [CHANGELOG.md](CHANGELOG.md) - [server.json](server.json): public MCP Registry metadata. ## Development ```bash uv sync --all-extras --dev uv run pytest uv run ruff check . uv run ruff format --check . uv run ty check ``` Licensed under [AGPL-3.0-or-later](LICENSE).