MCP Server (AI Assistant Integration)

MCP Server (AI Assistant Integration)

SceneItAll provides a Model Context Protocol (MCP) server that lets external AI assistants — such as Claude, ChatGPT, or any MCP-compatible client — connect to your projects and help you work.

Most tools are read-only, but the server also supports a set of mutation tools (scene updates, element tagging, availability changes, schedule edits). All mutations require explicit user confirmation before the AI assistant executes them.

What the AI assistant can do

Once connected, the AI assistant has access to 69 tools across twelve categories:

CategoryToolsExamples
Navigationlist_projects, set_project, set_episode, set_script, get_project_overview, get_context”Show me my projects”
Dashboardget_dashboard”How is the project looking overall?”
Scripts & sceneslist_episodes, list_scripts, list_scenes, get_scene_detail, update_scene*, bulk_update_scenes*, get_scenes_by_location, get_scenes_by_set, get_scene_comments”What exterior night scenes are in episode 3?”
Elements & breakdownlist_elements, get_element_detail, get_scene_breakdown, get_cast_list, get_element_usage, create_element*, update_element*, batch_move_element_category*, bulk_create_elements*“Rename the prop, then move all weapons into the stunts category”
Element taggingtag_element_to_scene*, bulk_tag_elements*, confirm_scene_element*, update_scene_element*, untag_element_from_scene*“Tag the red car to scene 12 — and untag the wrong one from scene 8”
Crew & peoplelist_crew, list_cast, search_people, get_availability, set_availability*, set_availability_range*, set_availability_bulk*“Who is available in March?”
Locationslist_locations, get_location_detail, get_scenes_at_location, create_location*“Which scenes are at the warehouse?”
Schedule & continuitylist_scenarios, compare_scenarios, get_schedule, get_one_liner, list_strips, list_schedule_days, get_day_summary, get_change_markers, get_call_sheet, get_full_fat_schedule, get_dood, get_element_dood, get_continuity_warnings”Any continuity issues in the latest script?”
Schedule mutationsmove_strips*, reorder_strips*, create_banner*, create_schedule_day*, update_schedule_day*, create_scenario*, update_scenario*, create_strips_from_scenes*, update_strip*“Branch a ‘tight’ scenario from the current one and try moving day 3 onto day 2”
Schedule settingsget_schedule_settings, update_schedule_settings*“What is our minimum turnaround? Bump it from 10 to 11 hours.”
Conflictslist_conflicts, validate_schedule_change”Are there any scheduling conflicts?”
Search & notificationssearch_project, get_notifications”Find scenes mentioning the red car”

* = requires user confirmation before execution

Role prompts

The MCP server includes pre-built role prompts that configure the AI assistant with deep production knowledge for a specific job. Invoke a prompt to switch the AI into that role’s mindset — it will prioritise the right tools, flag role-specific issues, and communicate in that role’s style.

PromptRoleFocus
first_ad1st Assistant DirectorSchedule integrity, conflicts, turnaround, day planning
script_supervisorScript SupervisorContinuity, scene ordering, script days, time-of-day logic
line_producerLine Producer / UPMCost-aware scheduling, hold days, overtime, crew logistics
location_managerLocation ManagerPermits, location grouping, company moves, weather contingency
casting_directorCasting DirectorCast availability, hold days, actor deal awareness

How it works

Your AI client  ──SSE──►  MCP Server  ──HTTP──►  SceneItAll API

The MCP server is a thin layer between your AI assistant and the SceneItAll API. It translates MCP tool calls into API requests using your authentication token. All existing permission checks apply — a viewer gets read-only access, an admin gets admin-level data, exactly the same as the app.

Connecting your AI assistant

The easiest way to connect is from the SceneItAll app itself: open the account menu (click your avatar in the top bar) and choose Connect AI Assistant to get copy-pasteable setup instructions for Claude and ChatGPT.

Claude Web

  1. Go to claude.ai → Settings → Connectors → Add
  2. Name: SceneItAll
  3. Remote MCP server URL: https://mcp.sceneitall.pro/sse
  4. Click Add — you’ll be prompted to sign in with your SceneItAll email and password

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "sceneitall": {
      "url": "https://mcp.sceneitall.pro/sse",
      "headers": {
        "Authorization": "Bearer YOUR_ACCESS_TOKEN"
      }
    }
  }
}

Replace YOUR_ACCESS_TOKEN with a valid Clerk session JWT. SceneItAll authenticates every request through Clerk; the easiest way to grab a token is to sign in to the SceneItAll web app, open DevTools, and copy the Authorization: Bearer … value from any outgoing API request. Tokens expire (typically every few minutes) — if Claude Desktop starts returning 401, copy a fresh one from the running web session. For long-running automation, use Clerk’s machine-to-machine API token flow instead of a user session.

Claude Code

claude mcp add sceneitall --transport sse \
  --url https://mcp.sceneitall.pro/sse \
  --header "Authorization: Bearer YOUR_ACCESS_TOKEN"

ChatGPT (Custom GPT)

  1. Go to ChatGPT → My GPTs → Create a GPT
  2. Click Create new action
  3. Set Authentication to OAuth:
    • Client ID: chatgpt-sceneitall
    • Client Secret: (ask your admin)
    • Authorization URL: https://mcp.sceneitall.pro/authorize
    • Token URL: https://mcp.sceneitall.pro/token
    • Scope: sceneitall:read
    • Token Exchange: POST
  4. Import schema from URL: https://mcp.sceneitall.pro/openapi.json
  5. Save and test

Security

  • Your permissions apply. The MCP server uses your token, so it can only access projects you’re a member of, at your role level.
  • Mutations require confirmation. The AI assistant must show you what it intends to change and wait for your approval before executing any write operation.
  • No personal data exposed. Contact details, addresses, agency info, measurements, and financial data are never returned by MCP tools.
  • Tokens stay in memory. Nothing is written to disk or logged.
  • Rate limited. The same API rate limits apply to MCP requests.

Typical workflow

  1. The AI assistant calls list_projects. If you have one project it is selected automatically; if multiple, the AI asks which one.
  2. Same for episodes (list_episodes) and scripts (list_scripts) — auto-selected if only one, otherwise the AI asks.
  3. From there, ask anything: scene breakdowns, crew availability, continuity warnings, schedule details, etc.
  4. Optionally, invoke a role prompt (e.g. first_ad) to configure the AI as a specialist for your department.
  5. If you ask the AI to make changes (tag elements, update availability, move strips), it will show you the proposed change and ask for confirmation first.