New Toolkit Request: Unstructured Transform (document parsing for AI)

[Jared Sulzdorf (jsulz)]

What it is

Unstructured Transform turns any document into structured, AI-ready data (partitioned, enriched, chunked, embedded). It handles 60+ file types (PDFs, Office docs, emails, images, scanned/handwritten files) through a single API and is used heavily in enterprise RAG and agent pipelines. A Composio toolkit would let agents submit documents for parsing, poll job status, and pull structured output without hand-rolling the REST client.

We are Unstructured (the provider). We maintain the API and are offering a ready-to-import OpenAPI spec + integrations.yaml to make this a low-lift build on your side. Happy to adjust to your current toolkit format.

Resources

• Website: https://unstructured.io/
• Product docs: https://docs.unstructured.io/transform/overview
• Base URL: https://platform-api.transform.unstructured.io/api/v1
• Get an API key: https://transform.unstructured.io/

Authentication

API key, enforced per request via the unstructured-api-key HTTP header (a custom header name, not Authorization/Bearer). Users copy their key from the Transform dashboard ("Get started" → "Get your API key"). Note: the key is declared as a per-operation header parameter in the OpenAPI spec rather than a top-level securityScheme, so the integrations.yaml below pins it explicitly.

Endpoints / proposed actions

The toolkit is the Jobs group — submit a document, track the job, retrieve structured output. These seven actions cover the full document-parsing flow.

Method	Path	Proposed action	Description
POST	/jobs/	UNSTRUCTURED_TRANSFORM_CREATE_JOB	Create a parsing job (multipart: input_files[] + request_data JSON of parse options)
GET	/jobs/	UNSTRUCTURED_TRANSFORM_LIST_JOBS	List jobs (filter by workflow_id, status)
GET	/jobs/{job_id}	UNSTRUCTURED_TRANSFORM_GET_JOB	Get job status/details
GET	/jobs/{job_id}/details	UNSTRUCTURED_TRANSFORM_GET_JOB_DETAILS	Get job processing details
GET	/jobs/{job_id}/download	UNSTRUCTURED_TRANSFORM_DOWNLOAD_JOB_OUTPUT	Download parsed output (file_id, optional node_id)
GET	/jobs/{job_id}/failed-files	UNSTRUCTURED_TRANSFORM_GET_JOB_FAILED_FILES	List files that failed in a job
POST	/jobs/{job_id}/cancel	UNSTRUCTURED_TRANSFORM_CANCEL_JOB	Cancel a running job

integrations.yaml

name: Unstructured Transform
description: Turn any document into structured, AI-ready data (partition, enrich, chunk, embed) via the Unstructured Platform API.
logo: https://unstructured.io/favicon.ico
categories:
  - document-processing
  - data-extraction
  - rag
get_current_user_endpoint: "/jobs/?limit=1"
auth_schemes:
  - scheme_name: unstructured_transform_api_key
    auth_mode: API_KEY
    proxy:
      base_url: "{{base_url}}"
      headers:
        unstructured-api-key: "{{api_key}}"
    fields:
      - name: api_key
        expected_from_customer: true
        displayName: Unstructured API Key
        description: "Your Unstructured API key. Get it from https://transform.unstructured.io ('Get started' > 'Get your API key')."
        type: string
        required: true
      - name: base_url
        expected_from_customer: false
        displayName: Base URL
        description: "Unstructured Platform API base URL."
        type: string
        required: false
        default: "https://platform-api.transform.unstructured.io/api/v1"

Notes on the config:

• Auth is a single API key sent as the unstructured-api-key header; no OAuth, no token refresh.
• base_url defaults to the hosted Platform API and is overridable for customers on dedicated/self-hosted deployments.
• A lightweight authenticated GET /jobs/ (list) works as a connection/health check if you validate keys at connect time.

The integrations.yaml fields above follow Composio's API_KEY integration format as we understand it; glad to conform to your current schema or provide anything else (Postman collection, sample payloads, test key).

Thanks!