DocIntel MCP server

What it does: Returns the enterprise contract version, lifecycle capabilities, and versioned workflow definitions.

Use it when: A client first connects and needs to discover supported governance, event, artifact, versioning, evaluation, and OAuth capabilities.

mcp_tool get_enterprise_capabilities '{}' | tool_data | jq '.'

What it does: Lists supported healthcare, finance, talent, employee mobility, and lease workflows with required inputs and packet support.

Use it when: Building a dynamic integration that should not hard-code available vertical workflows.

mcp_tool list_vertical_workflows '{}' | tool_data | jq '.'

What it does: Returns one workflow's versioned input, review, and packet contract.

Use it when: Preparing a form, validating an integration, or determining which documents a workflow requires.

mcp_tool get_workflow_schema '{"workflow":"healthcare_prior_auth"}' | tool_data | jq '.'

What it does: Checks input presence without starting or charging for a workflow run.

Use it when: Implementing preflight validation before workflow execution.

mcp_tool validate_workflow_inputs '{"workflow":"healthcare_prior_auth","inputs":{"document_ids":["YOUR_DOCUMENT_ID"],"policy_document_ids":["YOUR_POLICY_DOCUMENT_ID"]}}' | tool_data | jq '.'

What it does: Lists workspaces accessible to the authenticated user.

Use it when: Selecting tenant scope before listing, uploading, searching, or running workflows.

mcp_tool list_workspaces '{}' | tool_data | jq '.'

What it does: Lists personal documents or documents in one workspace.

Use it when: Discovering IDs, status, classification, language, or file metadata before downstream processing.

mcp_tool list_documents '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Returns metadata for one accessible document.

Use it when: Verifying classification, storage status, chunk count, embedding state, or source metadata.

mcp_tool get_document '{"document_id":"YOUR_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Creates a document record and short-lived signed cloud PUT URL so bytes bypass MCP and the backend proxy.

Use it when: Uploading documents safely, especially files too large for a JSON-RPC request.

mcp_tool create_document_upload '{"filename":"policy.pdf","content_type":"application/pdf","file_size":123456,"workspace_id":"YOUR_WORKSPACE_ID","redact_pii":false}' | tool_data | jq '.'

What it does: Verifies the uploaded cloud object and starts chunking.

Use it when: Completing the second step of direct document ingestion.

mcp_tool complete_document_upload '{"doc_id":"YOUR_NEW_DOCUMENT_ID","filename":"policy.pdf","content_type":"application/pdf","file_size":123456,"gcs_source_path":"RETURNED_GCS_SOURCE_PATH","workspace_id":"YOUR_WORKSPACE_ID","redact_pii":false}' | tool_data | jq '.'

What it does: Reports chunking/embedding status, counts, progress, and errors.

Use it when: Polling an asynchronous upload until it becomes chunked or embedded, or diagnosing a stalled ingestion.

mcp_tool get_ingestion_status '{"document_id":"YOUR_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Returns a document's chunk manifest.

Use it when: Auditing chunk boundaries, selecting chunk indices for a focused summary, or validating ingestion quality.

mcp_tool get_document_chunks '{"document_id":"YOUR_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Starts embedding after chunking has completed.

Use it when: Making a document available to semantic and hybrid retrieval.

mcp_tool embed_document '{"document_id":"YOUR_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Permanently deletes the document, source object, chunks, vectors, and dependent records.

Use it when: Withdrawing content, correcting an upload, or enforcing data lifecycle requirements. This is destructive and requires confirmation.

mcp_tool delete_document '{"document_id":"YOUR_DOCUMENT_ID","confirm":true}' | tool_data | jq '.'

What it does: Creates a video document and signed direct-upload URL.

Use it when: Uploading large videos without routing gigabytes through MCP or the application backend.

mcp_tool create_video_upload '{"filename":"training.mp4","content_type":"video/mp4","file_size":1073741824,"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Verifies the uploaded video and optionally begins transcript, frame, segment, and embedding processing.

Use it when: Completing direct video upload, with processing choices known at upload time.

mcp_tool complete_video_upload '{"doc_id":"YOUR_VIDEO_DOCUMENT_ID","filename":"training.mp4","content_type":"video/mp4","file_size":1073741824,"gcs_source_path":"RETURNED_GCS_SOURCE_PATH","workspace_id":"YOUR_WORKSPACE_ID","process_after_upload":true,"rights_confirmed":true,"transcript_language":"auto","max_frames":12,"segment_seconds":60,"embed_after_processing":true}' | tool_data | jq '.'

What it does: Lists accessible videos and current processing progress.

Use it when: Building a video library, selecting a video, or monitoring workspace video ingestion.

mcp_tool list_videos '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Starts metadata probing, audio transcription, frame sampling, timeline segmentation, and optional embedding.

Use it when: A video was uploaded without processing or must be reprocessed with a selected transcript language or segmentation strategy.

mcp_tool process_video '{"document_id":"YOUR_VIDEO_DOCUMENT_ID","rights_confirmed":true,"transcript_language":"hi-IN","max_frames":12,"segment_seconds":60,"embed_after_processing":true}' | tool_data | jq '.'

What it does: Returns current stage, percentage, timestamps, metadata, and processing errors.

Use it when: Showing progress or determining where a long video pipeline stalled.

mcp_tool get_video_status '{"document_id":"YOUR_VIDEO_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Returns timestamped segments and sampled frames.

Use it when: Building timeline navigation, chapter views, highlight review, or evidence linking.

mcp_tool get_video_timeline '{"document_id":"YOUR_VIDEO_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Returns timestamped transcript entries.

Use it when: Reviewing speech content, creating subtitles, searching spoken statements, or citing the exact time of a discussion.

mcp_tool get_video_transcript '{"document_id":"YOUR_VIDEO_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Returns sampled-frame metadata, timestamps, captions, and OCR text.

Use it when: Visual evidence matters in addition to speech, such as sports, inspections, training, or cultural performances.

mcp_tool get_video_frames '{"document_id":"YOUR_VIDEO_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Generates a short-lived view URL for one sampled frame.

Use it when: Displaying a cited frame securely without making cloud storage public.

mcp_tool get_video_frame_url '{"document_id":"YOUR_VIDEO_DOCUMENT_ID","frame_index":0}' | tool_data | jq '.'

What it does: Answers a question against embedded, timestamp-aware video evidence.

Use it when: Asking what happened, when it happened, what was said, or what was visible in a processed video.

mcp_tool search_video '{"document_id":"YOUR_VIDEO_DOCUMENT_ID","question":"What is discussed between 1:00 and 3:00?","limit":8}' | tool_data | jq '.'

What it does: Runs grounded hybrid retrieval, reranking, prompt assembly, and cited response generation over selected documents.

Use it when: Asking factual questions that must be grounded in an explicit workspace/document context.

mcp_tool search_knowledgebase '{"question":"What are the major risks and next actions?","document_ids":["YOUR_DOCUMENT_ID"],"workspace_id":"YOUR_WORKSPACE_ID","history":[],"redact_pii":false}' | tool_data | jq '.'

What it does: Queries an explicit authorized document set across multiple workspaces while preserving normalized citations.

Use it when: Relevant evidence spans projects or workspaces and the caller already knows which documents should participate.

mcp_tool search_federated_knowledgebase '{"question":"Compare the obligations across these records.","document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"history":[],"redact_pii":false}' | tool_data | jq '.'

What it does: Generates `executive`, `detailed`, `bullets`, `sections`, or custom summaries for one document.

Use it when: Producing a concise briefing, detailed review, structured outline, or targeted summary of selected chunks.

mcp_tool summarize_document '{"document_id":"YOUR_DOCUMENT_ID","summary_type":"executive","custom_prompt":"","chunk_indices":[],"redact_pii":false}' | tool_data | jq '.'

What it does: Generates one combined summary across multiple documents.

Use it when: Creating a case brief, portfolio summary, multi-policy digest, or cross-document narrative.

mcp_tool summarize_documents '{"document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"summary_type":"detailed","custom_prompt":"","redact_pii":false}' | tool_data | jq '.'

What it does: Returns similarity, unique points, and section-level differences between two documents.

Use it when: Comparing contract versions, current/prior policies, tax returns, candidate materials, or related reports.

mcp_tool compare_documents '{"document_id_1":"YOUR_DOCUMENT_ID_1","document_id_2":"YOUR_DOCUMENT_ID_2","redact_pii":false}' | tool_data | jq '.'

What it does: Creates a persistent conversation scoped to selected documents and optionally a workspace.

Use it when: A user needs multi-turn Q&A that can be resumed later.

mcp_tool create_chat_session '{"document_ids":["YOUR_DOCUMENT_ID"],"workspace_id":"YOUR_WORKSPACE_ID","title":"Policy review"}' | tool_data | jq '.'

What it does: Lists personal or workspace chat sessions.

Use it when: Building conversation history navigation or resuming previous work.

mcp_tool list_chat_sessions '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Returns a session, selected documents, and saved messages.

Use it when: Restoring conversation context before the next question.

mcp_tool get_chat_session '{"session_id":"YOUR_SESSION_ID"}' | tool_data | jq '.'

What it does: Renames a session or changes its selected documents.

Use it when: The subject evolves or the evidence set must be expanded or narrowed.

mcp_tool update_chat_session '{"session_id":"YOUR_SESSION_ID","title":"Updated policy review","document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"]}' | tool_data | jq '.'

What it does: Permanently removes a chat session.

Use it when: Cleaning up obsolete conversations or satisfying user data lifecycle requests. Requires confirmation.

mcp_tool delete_chat_session '{"session_id":"YOUR_SESSION_ID","confirm":true}' | tool_data | jq '.'

What it does: Performs grounded Q&A and optionally loads/saves persistent session history.

Use it when: Implementing normal DocIntel conversational behavior through MCP rather than a one-off retrieval call.

mcp_tool ask '{"question":"What changed since the previous discussion?","document_ids":["YOUR_DOCUMENT_ID"],"workspace_id":"YOUR_WORKSPACE_ID","session_id":"YOUR_SESSION_ID","history":[],"redact_pii":false}' | tool_data | jq '.'

What it does: Starts a governed conversation recording session in a workspace and language context.

Use it when: A client wants to collect speech or text turns and later publish the reviewed transcript as DocIntel knowledge.

mcp_tool start_conversation_recording '{"workspace_id":"YOUR_WORKSPACE_ID","language_code":"en-US"}' | tool_data | jq '.'

What it does: Records participant consent before turns are collected.

Use it when: A conversation must be explicitly consented before becoming a governed knowledge artifact.

mcp_tool confirm_conversation_consent '{"session_id":"YOUR_CONVERSATION_SESSION_ID","confirmed":true}' | tool_data | jq '.'

What it does: Adds one transcribed participant turn and returns the assistant's next prompt or collection state.

Use it when: A client has captured speech-to-text or typed input and needs the turn persisted immediately.

mcp_tool add_conversation_turn '{"session_id":"YOUR_CONVERSATION_SESSION_ID","transcript":"I am calling to provide an update about the account."}' | tool_data | jq '.'

What it does: Completes collection and prepares the transcript for review.

Use it when: The conversation is ready to move from guided collection into review and approval.

mcp_tool finish_conversation_recording '{"session_id":"YOUR_CONVERSATION_SESSION_ID"}' | tool_data | jq '.'

What it does: Returns recording status, editable transcript, document publication status, and review fields.

Use it when: A reviewer needs to inspect or edit the transcript before approval, or monitor publication after approval.

mcp_tool get_conversation_recording '{"session_id":"YOUR_CONVERSATION_SESSION_ID"}' | tool_data | jq '.'

What it does: Lists conversation recordings in an authorized workspace.

Use it when: Building review queues, operational dashboards, or conversation history for a workspace.

mcp_tool list_conversation_recordings '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Approves the reviewed transcript and publishes it as a document, chunks, embeddings, and searchable workspace knowledge.

Use it when: Human review is complete and the conversation should become available to grounded Q&A.

mcp_tool approve_conversation_transcript '{"session_id":"YOUR_CONVERSATION_SESSION_ID","transcript":"REVIEWED_TRANSCRIPT_TEXT","confirm":true}' | tool_data | jq '.'

What it does: Deletes the recording and related transcript document, chunks, vectors, turns, segments, and derived records.

Use it when: A workspace needs to withdraw an obsolete or incorrectly captured conversation. This requires explicit confirmation.

mcp_tool delete_conversation_recording '{"session_id":"YOUR_CONVERSATION_SESSION_ID","confirm":true}' | tool_data | jq '.'

What it provides: The conversation recording envelope, status, review data, publication status, and associated document context.

Helpful when: A resource-aware MCP client needs the full governed recording state as context.

mcp_request '{"jsonrpc":"2.0","id":201,"method":"resources/read","params":{"uri":"docintel://conversations/YOUR_CONVERSATION_SESSION_ID"}}'

What it provides: A compact transcript payload for review, editing, approval, or downstream prompt context.

Helpful when: The client wants transcript-only context without loading the full recording record.

mcp_request '{"jsonrpc":"2.0","id":202,"method":"resources/read","params":{"uri":"docintel://conversations/YOUR_CONVERSATION_SESSION_ID/transcript"}}'

What it does: Lists courses visible to the caller in one workspace.

Use it when: Building a student, teacher, or advisor course dashboard.

mcp_tool list_learning_courses '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Reads course metadata, curriculum, content, artifacts, questions, and authorized progress.

Use it when: Loading the complete current course experience.

mcp_tool get_learning_course '{"course_id":"YOUR_COURSE_ID"}' | tool_data | jq '.'

What it does: Creates a workspace-governed course with objectives and instructor context.

Use it when: A teacher or course administrator starts a learning experience.

mcp_tool create_learning_course '{"workspace_id":"YOUR_WORKSPACE_ID","title":"AI Foundations","course_code":"AI-101","semester":"Fall 2026","description":"Evidence-grounded AI learning","instructor_name":"Course Instructor","objectives":["Explain RAG","Evaluate grounded answers"]}' | tool_data | jq '.'

What it does: Updates course metadata or lifecycle status.

Use it when: Objectives, instructor, semester, description, or status changes.

mcp_tool update_learning_course '{"course_id":"YOUR_COURSE_ID","changes":{"description":"Updated course description","status":"active"}}' | tool_data | jq '.'

What it does: Deletes a course and its learning records while preserving source DocIntel assets.

Use it when: Removing a test or withdrawn course. Explicit confirmation is required.

mcp_tool delete_learning_course '{"course_id":"YOUR_COURSE_ID","confirm":true}' | tool_data | jq '.'

What it does: Saves the ordered module and lesson hierarchy.

Use it when: Designing or revising the course curriculum.

mcp_tool save_learning_curriculum '{"course_id":"YOUR_COURSE_ID","modules":[{"title":"RAG Foundations","description":"Retrieval and grounding","lessons":[{"title":"Hybrid Retrieval","description":"Keyword and vector retrieval"}]}]}' | tool_data | jq '.'

What it does: Enrolls an existing workspace user as student, teacher, advisor, or admin.

Use it when: Assigning a governed course persona.

mcp_tool enroll_learning_member '{"course_id":"YOUR_COURSE_ID","email":"student@example.com","persona":"student"}' | tool_data | jq '.'

What it does: Removes course enrollment without deleting the DocIntel user.

Use it when: A learner or staff member leaves the course.

mcp_tool remove_learning_member '{"course_id":"YOUR_COURSE_ID","user_id":"YOUR_USER_ID","confirm":true}' | tool_data | jq '.'

What it does: Maps an embedded document, recording, or video to a course, module, or lesson.

Use it when: Establishing the authoritative evidence scope for Tutor and Study Tools.

mcp_tool attach_learning_content '{"course_id":"YOUR_COURSE_ID","document_id":"YOUR_DOCUMENT_ID","module_id":"YOUR_MODULE_ID","lesson_id":"YOUR_LESSON_ID","title":"Hybrid Retrieval Lesson"}' | tool_data | jq '.'

What it does: Detaches content from a course without deleting the source asset.

Use it when: Correcting curriculum mapping or retiring course material.

mcp_tool remove_learning_content '{"course_id":"YOUR_COURSE_ID","asset_id":"YOUR_ASSET_ID","confirm":true}' | tool_data | jq '.'

What it does: Answers a learner question from the authorized course, module, or lesson evidence scope.

Use it when: A student needs grounded explanations with citations.

mcp_tool ask_learning_tutor '{"course_id":"YOUR_COURSE_ID","module_id":"YOUR_MODULE_ID","lesson_id":"YOUR_LESSON_ID","question":"How does hybrid retrieval improve grounding?","history":[],"response_language":"en"}' | tool_data | jq '.'

What it does: Generates and saves a summary, study guide, key concepts, flashcards, or interactive practice quiz.

Use it when: Turning scoped learning evidence into reusable study material.

mcp_tool generate_learning_artifact '{"course_id":"YOUR_COURSE_ID","artifact_type":"practice_questions","module_id":"YOUR_MODULE_ID","lesson_id":"YOUR_LESSON_ID","title":"Hybrid Retrieval Practice","custom_instruction":""}' | tool_data | jq '.'

What it does: Lists the caller's saved course study artifacts.

Use it when: Restoring generated study material across sessions.

mcp_tool list_learning_artifacts '{"course_id":"YOUR_COURSE_ID"}' | tool_data | jq '.'

What it does: Deletes one caller-owned study artifact.

Use it when: Regenerating or removing obsolete personal study material.

mcp_tool delete_learning_artifact '{"course_id":"YOUR_COURSE_ID","artifact_id":"YOUR_ARTIFACT_ID","confirm":true}' | tool_data | jq '.'

What it does: Grades and persists selected answers for a generated practice quiz.

Use it when: Providing immediate feedback, explanations, and resumable practice.

mcp_tool submit_learning_quiz '{"course_id":"YOUR_COURSE_ID","artifact_id":"YOUR_ARTIFACT_ID","answers":{"q1":["A","C"]},"replace":false}' | tool_data | jq '.'

What it does: Returns persistent quiz progress visible to the caller's course persona.

Use it when: Restoring learner state or reviewing permitted course progress.

mcp_tool get_learning_progress '{"course_id":"YOUR_COURSE_ID"}' | tool_data | jq '.'

What it does: Escalates a contextual question to a teacher or advisor.

Use it when: AI assistance should hand off to human judgment.

mcp_tool ask_learning_person '{"course_id":"YOUR_COURSE_ID","target_role":"teacher","question":"Can you review my retrieval design?","context":{"module":"RAG Foundations"}}' | tool_data | jq '.'

What it does: Lets an authorized teacher or advisor answer and update a learner question.

Use it when: Closing the human guidance loop with an auditable response.

mcp_tool answer_learning_question '{"course_id":"YOUR_COURSE_ID","question_id":"YOUR_QUESTION_ID","answer":"Review precision, recall, and citation quality.","status":"answered"}' | tool_data | jq '.'

What it does: Creates one durable job and signed upload URL per file manifest.

Use it when: Ingesting many documents with one monitored operation.

mcp_tool create_batch_upload '{"files":[{"filename":"one.pdf","content_type":"application/pdf","file_size":12345},{"filename":"two.pdf","content_type":"application/pdf","file_size":23456}],"workspace_id":"YOUR_WORKSPACE_ID","redact_pii":false,"idempotency_key":"batch-upload-001"}' | tool_data | jq '.'

What it does: Verifies uploaded objects and starts bounded parallel chunking.

Use it when: All or selected signed uploads from `create_batch_upload` have completed.

mcp_tool complete_batch_upload '{"batch_job_id":"YOUR_BATCH_JOB_ID","document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"concurrency":2}' | tool_data | jq '.'

What it does: Starts resumable parallel embedding with item-level outcomes.

Use it when: Many chunked documents must become searchable together.

mcp_tool start_batch_embedding '{"document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"workspace_id":"YOUR_WORKSPACE_ID","concurrency":2,"force":false,"idempotency_key":"batch-embed-001"}' | tool_data | jq '.'

What it does: Classifies documents in parallel with retryable item results.

Use it when: Organizing newly ingested content by type, domain, or workflow eligibility.

mcp_tool start_batch_classification '{"document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"workspace_id":"YOUR_WORKSPACE_ID","concurrency":2,"force":false,"idempotency_key":"batch-classify-001"}' | tool_data | jq '.'

What it does: Runs hierarchical map-reduce summarization across a large workspace/document set.

Use it when: A single model context cannot safely hold the entire corpus.

mcp_tool start_workspace_summary '{"workspace_id":"YOUR_WORKSPACE_ID","document_ids":["YOUR_DOCUMENT_ID_1","YOUR_DOCUMENT_ID_2"],"summary_type":"executive","custom_prompt":"Focus on risks and actions.","redact_pii":false,"language":"en","concurrency":2,"idempotency_key":"workspace-summary-001"}' | tool_data | jq '.'

What it does: Lists jobs with optional workspace, operation, and status filters.

Use it when: Building operations dashboards or finding a previous job.

mcp_tool list_batch_jobs '{"workspace_id":"YOUR_WORKSPACE_ID","operation":null,"status":null,"limit":25}' | tool_data | jq '.'

What it does: Returns stage, counters, attempts, progress, and errors.

Use it when: Polling an active job or diagnosing partial failure.

mcp_tool get_batch_status '{"batch_job_id":"YOUR_BATCH_JOB_ID"}' | tool_data | jq '.'

What it does: Returns aggregate and item-level outputs.

Use it when: Consuming successful results while understanding failed items.

mcp_tool get_batch_results '{"batch_job_id":"YOUR_BATCH_JOB_ID"}' | tool_data | jq '.'

What it does: Requeues only failed items rather than repeating successful work.

Use it when: Transient provider, quota, network, or extraction failures affected part of a batch.

mcp_tool retry_batch_failures '{"batch_job_id":"YOUR_BATCH_JOB_ID"}' | tool_data | jq '.'

What it does: Resumes a failed/interrupted batch from unfinished items.

Use it when: A worker, service, or deployment interruption stopped progress.

mcp_tool resume_batch_job '{"batch_job_id":"YOUR_BATCH_JOB_ID"}' | tool_data | jq '.'

What it does: Stops queued work while preserving completed outputs.

Use it when: A batch was configured incorrectly or no longer needs to run. Requires confirmation.

mcp_tool cancel_batch_job '{"batch_job_id":"YOUR_BATCH_JOB_ID","confirm":true}' | tool_data | jq '.'

What it does: Starts a supported domain workflow against accessible, processed documents.

Use it when: Moving from generic document intelligence into healthcare, finance, talent, mobility, or lease business processing.

mcp_tool start_vertical_workflow '{"workflow":"healthcare_clinical","document_ids":["YOUR_DOCUMENT_ID"],"workspace_id":"YOUR_WORKSPACE_ID","inputs":{}}' | tool_data | jq '.'

What it does: Returns workflow status, structured outputs, review packet, approval state, and errors.

Use it when: Polling a run or rendering its current business state.

mcp_tool get_vertical_run '{"vertical":"healthcare","run_id":"YOUR_RUN_ID"}' | tool_data | jq '.'

What it does: Lists accessible finance/tax or talent workflow runs.

Use it when: Building case history, reopening saved work, or filtering by status.

mcp_tool list_vertical_runs '{"vertical":"talent","workspace_id":"YOUR_WORKSPACE_ID","status":"all","limit":25}' | tool_data | jq '.'

What it does: Saves human-reviewed packet edits without approving them.

Use it when: A reviewer needs iterative editing, notes, and save/resume behavior before final approval.

mcp_tool save_vertical_review '{"vertical":"healthcare","run_id":"YOUR_RUN_ID","packet":{"review_status":"reviewed"},"notes":"Evidence reviewed.","persona":"clinical_reviewer"}' | tool_data | jq '.'

What it does: Applies the explicit human approval gate.

Use it when: An accountable reviewer confirms the workflow output may move forward. Requires `confirm:true`.

mcp_tool approve_vertical_run '{"vertical":"healthcare","run_id":"YOUR_RUN_ID","confirm":true,"packet":null,"notes":"Approved after review.","persona":"clinical_reviewer"}' | tool_data | jq '.'

What it does: Generates or ingests a review-ready PDF and returns document metadata or a signed URL.

Use it when: Producing a governed downloadable artifact after workflow and review completion.

mcp_tool generate_vertical_packet '{"vertical":"healthcare","run_id":"YOUR_RUN_ID","packet_type":"after_visit_summary","packet":null}' | tool_data | jq '.'

What it does: Reads durable lifecycle events after a monotonically increasing cursor.

Use it when: Reliably monitoring jobs, reviews, and workflows without losing events during reconnects.

mcp_tool list_operation_events '{"after_sequence":0,"resource_type":null,"resource_id":null,"limit":100}' | tool_data | jq '.'

What it does: Creates a filtered cursor subscription and optionally a signed HTTPS webhook.

Use it when: Integrating DocIntel lifecycle changes with enterprise orchestration, notifications, or downstream systems.

mcp_tool create_event_subscription '{"event_types":["batch.completed"],"workspace_id":"YOUR_WORKSPACE_ID","resource_type":"batch_job","resource_id":null,"webhook_url":"https://YOUR_HOST/docintel/events"}' | tool_data | jq '.'

What it does: Lists the caller's cursor and webhook subscriptions without exposing stored secrets.

Use it when: Auditing active integrations or finding a subscription ID.

mcp_tool list_event_subscriptions '{}' | tool_data | jq '.'

What it does: Deletes an owned subscription.

Use it when: Retiring an integration, rotating endpoints, or cleaning up a test webhook.

mcp_tool delete_event_subscription '{"subscription_id":"YOUR_SUBSCRIPTION_ID"}' | tool_data | jq '.'

What it does: Creates or retrieves a governed review task for a workflow run.

Use it when: Machine output must enter an accountable queue before approval or packet generation.

mcp_tool create_review_task '{"vertical":"healthcare","run_id":"YOUR_RUN_ID","title":"Review prior authorization packet","workspace_id":"YOUR_WORKSPACE_ID","priority":"high","metadata":{"case":"PA-1001"}}' | tool_data | jq '.'

What it does: Lists accessible pending, in-review, completed, rejected, or change-requested tasks.

Use it when: Building a reviewer work queue or checking decision status.

mcp_tool list_review_tasks '{"status":"pending"}' | tool_data | jq '.'

What it does: Assigns an accessible task to the current reviewer and marks it in review.

Use it when: A reviewer claims ownership before editing or deciding.

mcp_tool assign_review_task '{"task_id":"YOUR_REVIEW_TASK_ID"}' | tool_data | jq '.'

What it does: Records `approved`, `changes_requested`, or `rejected` with reviewer notes.

Use it when: Closing the human governance loop with an auditable outcome.

mcp_tool submit_review_decision '{"task_id":"YOUR_REVIEW_TASK_ID","decision":"approved","reviewer_notes":"Evidence and coding reviewed."}' | tool_data | jq '.'

What it does: Saves a summary, comparison, evidence map, report, or packet as reusable governed knowledge.

Use it when: Generated output should survive beyond one response and be reviewed, reused, or associated with source documents and traces.

mcp_tool save_knowledge_artifact '{"artifact_type":"reviewed_summary","title":"Approved risk summary","content":{"summary":"Reviewed content"},"workspace_id":"YOUR_WORKSPACE_ID","source_document_ids":["YOUR_DOCUMENT_ID"],"source_trace_id":"YOUR_TRACE_ID","status":"reviewed"}' | tool_data | jq '.'

What it does: Lists reusable artifacts in personal or workspace scope.

Use it when: Reopening reviewed outputs, building knowledge libraries, or locating reports generated earlier.

mcp_tool list_knowledge_artifacts '{"workspace_id":"YOUR_WORKSPACE_ID"}' | tool_data | jq '.'

What it does: Records document lineage, change summary, and caller-supplied changed pages.

Use it when: Tracking renewals, amendments, revised policies, updated tax records, or any replacement document.

mcp_tool register_document_version '{"document_id":"YOUR_NEW_DOCUMENT_ID","previous_document_id":"YOUR_OLD_DOCUMENT_ID","change_summary":"Renewal terms updated","changed_pages":[4,5]}' | tool_data | jq '.'

What it does: Lists lineage and change metadata for an accessible document family.

Use it when: Presenting history, selecting versions for comparison, or auditing what changed.

mcp_tool list_document_versions '{"document_id":"YOUR_NEW_DOCUMENT_ID"}' | tool_data | jq '.'

What it does: Lists requester-owned traces, optionally filtered by workspace.

Use it when: Finding recent RAG, MCP, workflow, tool, or model executions for investigation.

mcp_tool list_my_traces '{"workspace_id":"YOUR_WORKSPACE_ID","limit":50}' | tool_data | jq '.'

What it does: Returns requester-safe flow, timeline, model activity, spans, and correlated evaluations.

Use it when: Understanding how a response moved through authorization, retrieval, reranking, tools, prompts, model generation, and completion.

mcp_tool get_my_trace '{"trace_id":"YOUR_TRACE_ID"}' | tool_data | jq '.'

What it does: Correlates a deterministic quality baseline with an owned execution trace.

Use it when: Checking span success, response presence, and citation structure or routing weak runs to human evaluation.

mcp_tool evaluate_trace_quality '{"trace_id":"YOUR_TRACE_ID","evaluation_type":"groundedness"}' | tool_data | jq '.'