Document intelligence MCP server for extracting text, metadata, OCR output, structured data, embeddings, chunks, cache state, and supported-format information from PDFs, Office files, images, code, and many other formats.
Open the source and read safety notes before installing.
Source-backed facts for citing this resource, derived directly from the registry — also available as plain text for AI assistants.
Decision playbook
Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.
0
63
—
No baseline selected
No major trust-signal divergence detected in the current selection.
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Source provenance statusRequired
Marked as source-backed.
Metadata reviewed
No reviewed flag detected in metadata.
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Trust level risk gateRequired
Trust level does not block evaluation.
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Package verification flag
No package verification flag provided.
Checksum metadata
No checksum provided for downloaded artifact.
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
Baseline comparison available
No baseline peer selected yet.
Diverging trust signals identified
No major trust-signal divergence found.
Setup at a glance
Copy-ready — paste the snippet to get started.
Adoption plan
Current risk score 24/100. Use staged verification before broader rollout.
Validate source and review signals before any execution.
Confirm source provenanceRequired
Source URL/provenance metadata is present.
Confirm metadata review state
No review metadata found; increase manual validation.
Verify install payload
Install/config payload exists and can be inspected.
Confirm safety, privacy, and package integrity signals.
Review safety notesRequired
Safety notes are present.
Review privacy notesRequired
Privacy notes are present.
Verify package integrity metadata
No package verification/checksum metadata.
Adopt in controlled steps based on the selected plan.
Run in isolated sandbox firstRequired
Use a constrained sandbox and observe behavior across multiple tasks.
Roll out graduallyRequired
Roll out to a small cohort before wider usage.
Set monitoring and fallback
Define rollback path and monitor errors after adoption.
Evidence readiness
Missing required evidence: Metadata review. Risk score 31.
Source repository/provenance is listed.
Required in this preset
Review metadata is missing.
Required in this preset
Safety notes are present.
Required in this preset
Privacy notes are present.
Optional in this preset
Package integrity metadata is missing.
Optional in this preset
Install payload is available.
Required in this preset
Required gaps: Metadata review
Decision timeline
Blocking gaps: Check metadata review status. Risk 28.
triage
Source/provenance metadata is available.
triage
Review metadata is missing.
verify
Safety notes are available.
verify
Privacy notes are available.
verify
Package integrity metadata is missing.
rollout
Install payload is available.
Blockers: Check metadata review status
Prerequisite readiness
5 prerequisites to line up before setup.
Safety & privacy surface
5 safety and 5 privacy notes across 6 risk areas. Review closely: credentials & tokens, network access, third-party handling.
pip install "kreuzberg[all]"
kreuzberg mcpKreuzberg MCP Server exposes Kreuzberg's document intelligence engine through the Model Context Protocol. It lets Claude, Cursor, and custom MCP clients extract content from files, generate embeddings, chunk text, manage cache state, detect formats, and inspect supported extraction capabilities without writing custom extraction code.
The upstream MCP guide documents stdio mode with kreuzberg mcp as the default
local setup. Docker can also run the same MCP mode when file access and cache
volumes need to be controlled more explicitly.
These sources were reviewed on 2026-06-06. Prefer the live MCP integration guide, CLI usage docs, Docker guide, repository README, container package, and PyPI metadata for current install commands, server modes, feature flags, and runtime dependencies.
kreuzberg mcp.Install Kreuzberg with its optional feature bundle:
pip install "kreuzberg[all]"
Start the stdio MCP server:
kreuzberg mcp
Configure an MCP client with the documented command:
{
"mcpServers": {
"kreuzberg": {
"command": "kreuzberg",
"args": ["mcp"]
}
}
}
For containerized setups, mount only the document directories and config files the agent should access.
Kreuzberg can turn private files into model-visible text. That is useful, but it also means access boundaries matter. Mount or expose only approved directories, avoid broad workspace paths, and start with a small file before batch processing.
Review OCR, embedding, VLM, and LLM provider configuration before enabling features that can send document-derived content outside the local machine. Keep cache directories, logs, and extracted outputs out of shared repos unless they are approved for publication.
No kreuzberg-dev/kreuzberg entry, Kreuzberg MCP entry, or matching Kreuzberg
source URL was found in content/mcp.
Show that Kreuzberg MCP Server is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/mcp/kreuzberg-mcp-server)Kreuzberg MCP Server side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
| Field | Document intelligence MCP server for extracting text, metadata, OCR output, structured data, embeddings, chunks, cache state, and supported-format information from PDFs, Office files, images, code, and many other formats. Open dossier | MCP server that converts PDFs, Office files, images, audio, webpages, YouTube transcripts, and existing Markdown files into Markdown. Open dossier | MCP server for connecting Claude to AFFiNE Cloud or self-hosted AFFiNE workspaces, documents, databases, comments, collections, folders, tags, notifications, blobs, access tokens, semantic page composition, templates, edgeless canvas data, and workspace organization workflows over stdio or HTTP. Open dossier | Microsoft-maintained MCP server for converting HTTP, HTTPS, file, and data URIs into Markdown through the MarkItDown document-conversion library. Open dossier |
|---|---|---|---|---|
| Next steps | ||||
| Trust | ||||
| Review status | Not reviewed | Not reviewed | Not reviewed | Not reviewed |
| Package trust | Package not verified | Package not verified | Package not verified | Package not verified |
| Source provenance | Source-backed | Source-backed | Source-backed | Source-backed |
| Submitter | oktofeesh1 | oktofeesh1 | oktofeesh1 | oktofeesh1 |
| Install risk | Review first | Review first | Review first | Review first |
| Notes | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ |
| Brand | — | — | — | |
| Category | mcp | mcp | mcp | mcp |
| Source | Source-backed | Source-backed | Source-backed | Source-backed |
| Author | Kreuzberg | zcaceres | DAWNCR0W | Microsoft |
| Added | 2026-06-06 | 2026-06-05 | 2026-06-06 | 2026-06-06 |
| Platforms | ||||
| Harness | ||||
| Source repo | — | — | — | — |
| Safety notes | ✓Kreuzberg MCP can read local files supplied to extraction tools and can process batches of files when paths are provided. OCR, structured extraction, embeddings, and VLM features may invoke local or provider-hosted models depending on configuration. Cache tools can warm, inspect, or clear model/cache state; review cache directories and mounted volumes in shared environments. Docker deployments should mount only the directories the agent is allowed to read. Treat extracted text, metadata, structured fields, embeddings, and chunks as sensitive outputs when source documents are private. | ✓Markdownify MCP Server can read local files and convert documents, spreadsheets, presentations, images, audio files, and Markdown files. File conversion tools should be restricted with MD_ALLOWED_PATHS or container mounts before using the server with sensitive directories. Audio transcription, image OCR, document conversion, and webpage extraction can expose hidden text, metadata, comments, and embedded content. Webpage, YouTube, and Bing tools retrieve external content that can contain prompt-injection text or unsafe instructions. Review converted Markdown before using it as source material for decisions, publications, or further tool calls. | ✓The full tool surface can create, update, publish, revoke, move, replace, and delete AFFiNE workspaces, documents, folders, collections, comments, database rows, blobs, surface elements, tags, and access tokens. Start with AFFINE_TOOL_PROFILE=read_only or disable groups such as destructive, admin, blobs, users, access_tokens, docs.database, or write when the assistant only needs discovery or reading. HTTP mode exposes MCP endpoints and must be protected with bearer or OAuth authentication, HTTPS, allowed origins, and deployment-level access controls. AFFiNE Cloud requires API tokens for MCP usage; avoid trying to automate email/password login against Cloud deployments. Cookie and password-based auth can grant broad account access and should be avoided for unattended or shared deployments. | ✓MarkItDown MCP exposes one `convert_to_markdown(uri)` tool that accepts `http:`, `https:`, `file:`, and `data:` URIs. The server runs with the privileges of the user that starts it and can read any file that user can access when given a matching file URI. Streamable HTTP and SSE modes are local-use alternatives to stdio; the README warns not to bind them to non-localhost interfaces unless the security implications are understood. The HTTP and SSE server modes have no authentication, so any local process or user that can reach the bound interface can use the conversion tool. Use a container, virtual machine, or dedicated low-privilege user when converting untrusted documents or when file access must be tightly bounded. |
| Privacy notes | ✓Documents may contain PII, contracts, invoices, source code, screenshots, medical data, financial data, credentials, or proprietary design information. Extracted metadata can reveal filenames, authors, timestamps, document structure, attachments, image details, and software fingerprints. OCR and structured extraction can expose text that was not previously copyable from scanned PDFs or images. Embedding and VLM/LLM configuration may send document-derived content to external model providers if configured that way. Cache directories, logs, model downloads, and MCP transcripts may retain document-derived context. | ✓File paths, file contents, extracted text, OCR output, audio transcripts, webpage text, search results, prompts, tool arguments, and conversion errors may be visible to the MCP client and model provider. Documents and media can contain personal data, customer material, credentials, internal plans, meeting recordings, legal drafts, and proprietary screenshots. Avoid pointing the server at broad home, workspace, downloads, or cloud-sync directories unless every file in scope is approved for model access. | ✓Workspaces, document titles, document bodies, comments, tags, database schemas, database rows, edgeless canvas data, notifications, user profiles, access-token metadata, blobs, and exported markdown can be sent to the MCP client and model. API tokens, cookies, passwords, bearer tokens, OAuth config, and saved config files should be treated as secrets and kept out of prompts, logs, screenshots, issues, and committed MCP config. Generated or modified documents can persist in AFFiNE and may become visible to collaborators depending on workspace permissions and sharing state. Blob upload, cleanup, export, and publish/revoke workflows can expose files, generated content, or collaboration state beyond the current prompt. | ✓File paths, file contents, extracted Markdown, conversion errors, remote URLs, data URIs, and document metadata may become visible to the MCP client and model provider. Converting Office files, PDFs, media, web pages, or archives can reveal hidden text, comments, tracked changes, metadata, notes, image text, links, and embedded objects. Docker mounts make every mounted file readable to the MCP server path inside the container; mount only approved directories and prefer read-only mounts when possible. Remote URI conversion can reveal research targets, customer URLs, internal endpoints, or private document locations to upstream servers and to the model transcript. |
| Prerequisites |
|
|
|
|
| Install | | | | |
| Config | | | | |
| Citations | ||||
| Claim | Unclaimed | Unclaimed | Unclaimed | Unclaimed |
Loading live community signals…
A short, calm digest of reviewed Claude resources. Unsubscribe any time.