MCP endpoint (Streamable HTTP, one JSON-RPC message per POST)

Tools: get_suggested_workflow, start_task_task_creation, start_task_creation, list_task_creations, get_task_creation, rename_task_creation, add_task_creation_contribution, add_task_creation_evidence, remove_task_creation_evidence, create_task_creation_phone_link, revoke_task_creation_phone_link, edit_task_creation_contribution, retry_task_creation_contribution, select_task_creation_sections, renew_task_creation_link, start_task_creation_take, register_task_creation_track, upload_task_creation_chunk, finish_task_creation_take, edit_task_creation, approve_task_creation, accept_task_creation_recording, analyze_task_creation, save_task_creation_draft, save_task_creation_comments, submit_task_creation_context, revise_task_creation, publish_task_creation, retry_task_creation, discard_task_creation, list_my_resources, get_my_resource, preview_my_resource, download_my_resource, save_my_resource, rename_my_resource, delete_my_resource, retry_my_resource, start_resource_recording, list_task_resources, link_task_resource, unlink_task_resource, match_task_resources, freeze_task_resources, list_my_files, upload_my_file, complete_my_file, download_my_file, preview_my_file, rename_my_file, delete_my_file, list_task_files, set_task_file_access, set_task_file_visibility, upload_task_file, complete_task_file, download_task_file, preview_task_file, rename_task_file, delete_task_file, get_contact_notifications, set_contact_notifications, get_notification_preferences, set_notification_preferences, get_notifications, wait_for_updates, list_open_tasks, list_task_categories, list_task_posters, search_tasks, get_task, get_member, list_task_versions, list_task_applicants, list_messages, send_message, list_task_comments, comment_on_task, vote_on_comment, delete_comment, get_more_tools, submit_feature_request, list_feature_requests, get_safety_policy, report_content, list_my_content_reports, list_content_reports, resolve_content_report, restore_content, report_problem, list_problem_reports, set_up_account, get_my_profile, update_my_profile, update_my_photo, delete_my_account, suggest_tasks, complete_text, list_my_tasks, get_task_performance, get_task_draft, discard_task_draft, find_my_linkedin_profile, import_linkedin_profile, apply_to_task, list_my_applications, withdraw_application, list_task_applications, accept_application, decline_application, edit_task, close_task, contact_for_help, list_help_requests, get_task_work, submit_work, request_changes, approve_work, rate_poster, request_cancellation, cancel_work, list_connected_agents, disconnect_agent, invite_by_email, get_my_invite_link, list_invite_links, revoke_invite_link, list_waitlist, invite_from_waitlist, remind_about_work. Requires an OAuth access token; API keys are refused. Every tool accepts an optional conversation_id returned by an earlier tool call. Reuse it within the same conversation, including after reconnecting; pass "new" to start a new conversation. Optional context describes a non-sensitive goal, and llm_model identifies the calling model when known; both are analytics-only and removed before business validation. The MCP-only diagnostic tool get_more_tools requires context and reports a screened capability gap to PostHog when configured; it does not add tools or file a support request.