{
  "markdown": "# AlbumentationsX MCP\n\nModel Context Protocol server for [AlbumentationsX](https://github.com/albumentations-team/AlbumentationsX): inspect datasets, preview augmentations, refine them with visual feedback, and export reproducible pipelines.\n\n<!-- mcp-name: io.github.dKosarevsky/albu-mcp -->\n\n[![CI](https://github.com/dKosarevsky/albu-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/dKosarevsky/albu-mcp/actions/workflows/ci.yml)\n[![PyPI](https://img.shields.io/pypi/v/albumentationsx-mcp)](https://pypi.org/project/albumentationsx-mcp/)\n[![Python](https://img.shields.io/badge/python-3.10--3.13-blue)](pyproject.toml)\n[![MCP Registry](https://img.shields.io/badge/MCP%20Registry-active-green)](https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.dKosarevsky/albu-mcp)\n[![skills.sh](https://skills.sh/b/dKosarevsky/albu-mcp)](https://skills.sh/dKosarevsky/albu-mcp)\n\n![Baseline and adjusted AlbumentationsX preview contact sheets](https://raw.githubusercontent.com/dKosarevsky/albu-mcp/main/docs/assets/demo/comparison_contact_sheet.png)\n\nAsk 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.\n\n## Install\n\n### Claude Desktop\n\n[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.\n\n### Other MCP Hosts\n\nRun the published server with bounded local access:\n\n```bash\nuvx --from albumentationsx-mcp albumentationsx-mcp \\\n  --allowed-root /absolute/path/to/images \\\n  --artifact-root /absolute/path/to/albu-artifacts\n```\n\n`run_first_preview` requires the default `full` or `dataset` capability profile. The smaller `review` profile uses the\nexplicit validate/render fallback in the usage guide, or you can restart with `dataset` or `full`; see\n[configuration](docs/CONFIGURATION.md). Copyable host configurations are in [the install guide](docs/INSTALL.md).\nThe repository also contains a native Codex plugin bundle. `npx skills add dKosarevsky/albu-mcp` installs agent guidance, not the MCP server.\n\n## First Preview\n\nAfter connecting the server, ask your host:\n\n```text\nRun the host smoke check. If preview_ready is true, call run_first_preview for /absolute/path/to/images with low\nintensity and at most 8 images. Show me the contact sheet. When I mention a specific result, call\ntrace_preview_variant before adjusting it.\n```\n\n`run_host_smoke_check` returns `preview_ready` and a `preview_request_template`. If resource reads are unavailable, call\n`get_workflow_example` with `example_id=\"client-smoke\"`.\n\nTry the [classification robustness use case](docs/use-cases/CLASSIFICATION_ROBUSTNESS.md), or follow the\n[First 10 Minutes guide](docs/FIRST_10_MINUTES.md). The `validate_preview_request` fallback, batch previews, and how to\ncompare preview runs are in [Usage](docs/USAGE.md). Use `too_noisy:high` or `exposure_too_weak:medium`, then optionally\nshare one redacted loop through [first-preview feedback](docs/FIRST_PREVIEW_FEEDBACK.md).\nIf setup fails, read `albumentationsx://diagnostics/guide` and call `diagnose_environment` for bounded remediation actions.\n\n## Capabilities\n\n- Transform discovery, schemas, recipes, and pipeline validation.\n- Classification, detection, segmentation, OCR, bbox, mask, keypoint, and dataset-quality workflows.\n- Deterministic previews, contact sheets, annotation overlays, comparison, ranking, and reports.\n- Interactive MCP Apps review with a text-only fallback for other hosts.\n- Structured feedback, tuning sessions, and Python, JSON, or YAML export.\n- Runtime-aware CPU `torch.Tensor` pipeline validation and guarded Python handoff.\n- MCP `2026-07-28` plus legacy negotiation; stable agent workflow resources, diagnostics, and contract snapshots.\n\nThe 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`.\n\n## Integrations\n\n- [Official Albumentations MCP guide](https://albumentations.ai/docs/integrations/mcp/)\n- [Official MCP Registry entry](https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.dKosarevsky/albu-mcp)\n- [skills.sh agent skill](https://skills.sh/dKosarevsky/albu-mcp)\n- [Upstream documentation PR](https://github.com/albumentations-team/AlbumentationsX/pull/289)\n\n## Documentation\n\n- [Install and host configuration](docs/INSTALL.md)\n- [Runtime settings and capability profiles](docs/CONFIGURATION.md)\n- [First 10 minutes](docs/FIRST_10_MINUTES.md)\n- [First-preview feedback](docs/FIRST_PREVIEW_FEEDBACK.md)\n- [Usage](docs/USAGE.md) and [recipes](docs/RECIPES.md)\n- [CPU Tensor Compose validation and export](docs/TORCH_CPU_COMPOSE.md)\n- [MCP Apps review](docs/MCP_APPS_REVIEW.md) and [compatibility policy](docs/COMPATIBILITY.md)\n- [Documentation index](docs/INDEX.md)\n- [CHANGELOG.md](CHANGELOG.md)\n- [server.json](server.json): public MCP Registry metadata.\n\n## Development\n\n```bash\nuv sync --all-extras --dev\nuv run pytest\nuv run ruff check .\nuv run ruff format --check .\nuv run ty check\n```\n\nLicensed under [AGPL-3.0-or-later](LICENSE).\n",
  "bytes": 5309,
  "sha": "4a3916245c00c8feedc6f0022de4afef73fd0667dfd5e1d57257e86b283501a2",
  "repo_slug": "dkosarevsky/albu-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_dkosarevsky_albu_mcp_da926496/readme"
}