feat: add Unstract MCP integration for document processing #185
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change | ||||
|---|---|---|---|---|---|---|
| @@ -0,0 +1,130 @@ | ||||||
| --- | ||||||
| description: Unstract - LLM-powered document data extraction via MCP | ||||||
| mode: subagent | ||||||
| tools: | ||||||
| read: true | ||||||
| write: false | ||||||
| edit: false | ||||||
| bash: true | ||||||
| glob: false | ||||||
| grep: false | ||||||
| webfetch: true | ||||||
| unstract_tool: true | ||||||
| mcp: | ||||||
| unstract: true | ||||||
| --- | ||||||
|
|
||||||
| # Unstract - Document Processing | ||||||
|
|
||||||
| <!-- AI-CONTEXT-START --> | ||||||
|
|
||||||
| ## Quick Reference | ||||||
|
|
||||||
| - **Purpose**: Extract structured data from unstructured documents (PDFs, images, DOCX, etc.) | ||||||
| - **MCP Server**: `unstract/mcp-server` (Docker) or `@unstract/mcp-server` (npx) | ||||||
| - **Tool**: `unstract_tool` - submits files to Unstract API, polls for completion, returns structured JSON | ||||||
| - **Credentials**: `UNSTRACT_API_KEY` + `API_BASE_URL` in `~/.config/aidevops/mcp-env.sh` | ||||||
|
There was a problem hiding this comment. There's an inconsistency in the environment variable name for the Unstract API base URL. This line mentions
Suggested change
|
||||||
| - **Docs**: https://docs.unstract.com/unstract/unstract_platform/mcp/unstract_platform_mcp_server/ | ||||||
|
coderabbitai[bot] marked this conversation as resolved.
|
||||||
| - **GitHub**: https://github.com/Zipstack/unstract | ||||||
|
|
||||||
| **On-demand loading**: This MCP is disabled globally and enabled per-agent when document extraction is needed. | ||||||
|
|
||||||
| <!-- AI-CONTEXT-END --> | ||||||
|
|
||||||
| ## What is Unstract? | ||||||
|
|
||||||
| Unstract is a no-code LLM platform that structures unstructured documents. It provides: | ||||||
|
|
||||||
| - **Prompt Studio**: Visual environment to define extraction schemas | ||||||
| - **API Deployments**: Turn any document into JSON via REST API | ||||||
| - **ETL Pipelines**: Batch process documents into databases | ||||||
| - **MCP Server**: Integrate extraction into AI agent workflows | ||||||
|
|
||||||
| ## Supported File Types | ||||||
|
|
||||||
| | Category | Formats | | ||||||
| |----------|---------| | ||||||
| | Documents | PDF, DOCX, DOC, ODT, TXT, CSV, JSON | | ||||||
| | Spreadsheets | XLSX, XLS, ODS | | ||||||
| | Presentations | PPTX, PPT, ODP | | ||||||
| | Images | PNG, JPG, JPEG, TIFF, BMP, GIF, WEBP | | ||||||
|
|
||||||
| ## MCP Tool | ||||||
|
|
||||||
| ### `unstract_tool` | ||||||
|
|
||||||
| Submits a file to the Unstract API, polls for completion, and returns structured extraction results. | ||||||
|
|
||||||
| **Parameters**: | ||||||
| - `file_path` (required): Path to the document to process | ||||||
| - `include_metadata` (optional): Include extraction metadata in response | ||||||
| - `include_metrics` (optional): Include processing metrics (tokens, cost) | ||||||
|
|
||||||
| **Example prompt**: "Process the document /tmp/invoice.pdf" | ||||||
|
|
||||||
| ## Setup | ||||||
|
|
||||||
| ### 1. Get API Credentials | ||||||
|
|
||||||
| 1. Sign up at https://unstract.com/start-for-free/ (14-day free trial) | ||||||
| 2. Create a Prompt Studio project and define your extraction schema | ||||||
| 3. Deploy as an API endpoint | ||||||
| 4. Copy the API key and deployment URL | ||||||
|
|
||||||
| ### 2. Store Credentials | ||||||
|
|
||||||
| ```bash | ||||||
| # Add to ~/.config/aidevops/mcp-env.sh: | ||||||
| export UNSTRACT_API_KEY="your_api_key_here" | ||||||
| export UNSTRACT_API_BASE_URL="https://us-central.unstract.com/deployment/api/your-deployment-id/" | ||||||
| ``` | ||||||
|
|
||||||
| ### 3. OpenCode Configuration (On-Demand) | ||||||
|
|
||||||
| The MCP is configured in OpenCode but disabled globally. It loads on-demand when this subagent is invoked. | ||||||
|
|
||||||
| See `configs/mcp-templates/unstract.json` for the configuration template. | ||||||
|
|
||||||
| ### 4. Claude Desktop Configuration (Docker) | ||||||
|
|
||||||
| ```json | ||||||
| { | ||||||
| "mcpServers": { | ||||||
| "unstract_tool": { | ||||||
| "command": "/usr/local/bin/docker", | ||||||
| "args": [ | ||||||
| "run", "-i", "--rm", | ||||||
| "-v", "/tmp:/tmp", | ||||||
| "-e", "UNSTRACT_API_KEY", | ||||||
| "-e", "API_BASE_URL", | ||||||
| "unstract/mcp-server", | ||||||
| "unstract" | ||||||
| ], | ||||||
| "env": { | ||||||
| "UNSTRACT_API_KEY": "", | ||||||
|
There was a problem hiding this comment. The placeholder for
Suggested change
|
||||||
| "API_BASE_URL": "https://us-central.unstract.com/deployment/api/.../" | ||||||
| } | ||||||
| } | ||||||
| } | ||||||
| } | ||||||
| ``` | ||||||
|
|
||||||
| ## Use Cases | ||||||
|
|
||||||
| - **Invoice processing**: Extract line items, totals, vendor info from invoices | ||||||
| - **Bank statement parsing**: Structure transaction data from varied bank formats | ||||||
| - **Insurance claims**: Extract claim details from forms and supporting documents | ||||||
| - **KYC/onboarding**: Parse identity documents and application forms | ||||||
| - **Contract analysis**: Extract key terms, dates, parties from legal documents | ||||||
|
|
||||||
| ## Integration with aidevops | ||||||
|
|
||||||
| This subagent is referenced by agents that need document extraction capabilities. The MCP loads only when document processing tasks are detected. | ||||||
|
|
||||||
| **Trigger keywords**: document, extract, parse, invoice, statement, PDF, OCR, unstructured | ||||||
|
|
||||||
| ## Related | ||||||
|
|
||||||
| - `tools/context/mcp-discovery.md` - On-demand MCP loading pattern | ||||||
| - `.agent/aidevops/mcp-integrations.md` - All MCP integrations | ||||||
| - `configs/mcp-templates/unstract.json` - OpenCode config template | ||||||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,12 @@ | ||
| { | ||
| "unstract": { | ||
| "type": "local", | ||
| "command": [ | ||
| "/bin/bash", | ||
| "-c", | ||
| "source ~/.config/aidevops/mcp-env.sh && docker run -i --rm -v /tmp:/tmp -e UNSTRACT_API_KEY -e API_BASE_URL=\"$UNSTRACT_API_BASE_URL\" unstract/mcp-server unstract" | ||
| ], | ||
| "enabled": false, | ||
| "_comment": "Disabled globally. Enabled on-demand via services/document-processing/unstract.md subagent. Requires UNSTRACT_API_KEY and UNSTRACT_API_BASE_URL in ~/.config/aidevops/mcp-env.sh" | ||
| } | ||
| } |
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.