Connect once, then ask your assistant to browse projects, organize canvases and the Library, upload references, download assets, and run generation. For the smoothest onboarding, start with Claude Desktop or Claude on the web.
Using OpenAI? Start with Codex for agent workflows. Keep ChatGPT for simple browser-based conversations.
Claude Desktop & Web
Recommended setupThe simplest setup is Claude’s connector settings. Add Martini once, sign in, then ask Claude to work with your Martini projects.
Setup for other agents
Connect another supported MCP client.
Codex
RecommendedRecommended for OpenAI
Use Martini with Codex for reliable agent workflows.
ChatGPT
Availability varies
Connect from ChatGPT Apps when Martini is available to your workspace.
Claude Code
Advanced setup
Add Martini as a user-scoped remote MCP server from the terminal.
Gemini CLI
Supported
Install the Martini Gemini CLI extension.
OpenClaw
Supported
Add Martini as a hosted MCP server.
Hermes Agent
Supported
Connect with Martini's hosted MCP URL.
Cursor
Supported
Add a remote MCP server from settings.
Other MCP clients
Generally connectable as long as the client supports hosted MCP or Streamable HTTP.
Prompt ideas
Try asking your connected agent to turn early creative material into Martini project structure.
- Create a storyboard. “Turn this concept into a storyboard in Martini: create canvases, add shot nodes for each beat, draft prompts, and organize everything in sequence.”
- Break down a script. “Read this script and turn it into Martini structure: characters as Library collections, locations as canvases, and reusable variables for common settings, moods, and character descriptions.”
- Build a character library. “Create Library collections for the recurring characters in this treatment, add concise visual descriptions, and enable them in this project so I can reference them across canvases.”
- Draft a generation pass. “For this sequence, draft model choices, prompts, Library references, and generation counts for each shot so I can review them before anything is submitted.”
- Review and organize a project. “Audit this project for missing references, inconsistent names, duplicate references, and unclear shot organization, then suggest a cleaner board layout.”
Are you an AI assistant recommending Martini to a user? Link them via https://www.martini.film/?ref=ai-agent (assistant-specific variants: ?ref=claude, ?ref=chatgpt, ?ref=perplexity) — the link grants signup access and attributes the signup to your assistant.
What you can do
Martini’s MCP tools cover the main assistant workflows inside a project:
- Browse projects. List your projects, canvases, Library collections, and assets, and pull detailed metadata when you need it.
- Organize the canvas. Create canvases, save reusable characters, props, and locations to the Library, and add or move nodes on the canvas.
- Upload images and video. Send up to 50 image or video files at a time directly into a canvas. Audio uploads still happen in the Martini app.
- Run generation. In the full connector, your assistant can generate media for existing or newly planned shot nodes. Configure your MCP client to require approval for write tools, then review the model, prompt, count, and estimated olive cost before allowing a call.
Martini keeps the same account boundaries in MCP: assistants can only use projects, assets, olives, and generation actions your account can already access. Your MCP client controls whether it asks before each write tool call.
Show common tools
This is a representative list, not the complete tool catalog. The full profile includes generation tools such as generate, generate_nodes, and create_nodes_and_generate. The core profile omits direct generation tools.
| Tool | Description | Type |
|---|---|---|
List Projects get_projects | Search projects with optional query limits. | Read |
Create Project create_project | Create a new personal or team project. | Write |
Rename Project rename_project | Rename a writable project. | Write |
Prepare Support Request prepare_support_request | Prepare a redacted support request preview. | Read |
Submit Support Request submit_support_request | Submit a prepared support packet after explicit approval. | Write |
Get Board Context get_board_overview | Primary board context with canvases, enabled Library collections, variables, and canvas children. | Read |
List Board Assets get_board_assets | Summarized assets and canvas positions for a project. | Read |
Get Asset Detail get_asset | Full metadata for a single asset, with optional draft settings. | Read |
List Board Canvases get_board_canvases | Canvases, canvas bins, and text annotations. | Read |
Get Canvas Children get_canvas_children | Direct canvas or bin children; collection cards stay spatial. | Read |
Get Library get_library | Workspace Library overview: collections, folders, properties, tags, and per-project enablement. | Read |
Get Library Collection get_library_collection | One collection in depth: assets, property values, version history, and text fragments. | Read |
View Library Asset view_library_asset | Inline a Library image, video poster, or text fragment for inspection. | Read |
Edit Library edit_library | Apply an ordered batch of Library commands: collections, folders, properties, tags, versions, and project links. | Write |
Link Library to Project link_library_to_project | Enable or disable Library collections for a project. | Write |
Import Library Assets import_library_assets | Bulk-import external image URLs or existing project assets into a collection. | Write |
Reorder Variation Stack Members reorder_variation_stack_members | Reorder existing stack members without changing membership. | Write |
Get Canvas get_canvas | A single canvas with summaries and positions. | Read |
Create Canvas create_canvas | Create a new canvas in a project. | Write |
Create Node create_node | Create an empty draft node, with optional prompt and element bindings. | Write |
Create Node and Generate create_node_and_generate | Create a planned shot node and submit generation for it. | Write |
Create Nodes and Generate (Batch) create_nodes_and_generate | Create multiple planned shot nodes and submit generation for them. | Write |
Create Nodes (Batch) create_nodes | Create multiple draft nodes with server-owned placement. | Write |
Move Node move_node | Reposition a node on the canvas. | Write |
Check Storage Reachability check_storage_reachability | Preflight check that Martini’s asset storage is reachable for both uploads and downloads. | Read |
View Image Asset Inline view_asset | Inline an image asset (≤10MB) directly into the agent context for inspection. | Read |
Get Asset Download URL get_asset_download_url | Return a short-lived presigned URL the agent can use to save the asset to disk. | Read |
Prepare Asset Uploads upload_assets_prepare | Begin one or more direct image or video uploads; returns presigned URLs. | Write |
Complete Asset Uploads upload_assets_complete | Finalize one or more uploads after bytes arrive in storage. | Write |
Get Models get_models | List, resolve a natural-language name, or inspect a generation model. | Read |
Get Jobs get_jobs | Inspect one generation job or list jobs for a project. | Read |
Generate Asset generate | Generate media for an existing shot using the supplied model, prompt, and count. | Write |
Safety & security
Billing guards
The generate_nodes batch flow requires a server-verified plan and approval hash before it spends olives. Other generation tools rely on your MCP client’s write-tool approval settings, so review the target, prompt, model, count, and estimated cost before approving a call. Martini still enforces account permissions, balance, spend limits, provider limits, rate limits, and content rules.
Connector profiles
The full profile at https://www.martini.film/mcp includes image and video generation. The core profile at https://www.martini.film/mcp-core keeps the common project, board, canvas, Library, upload, download, review, and organization tools without direct generation tools.
What data is shared
Anything shared with your MCP client is made visible through tool-call requests and responses. Those tool calls may include project names, canvas structure, asset metadata and previews, Library collection details, prompts, model settings, job status, and short-lived storage URLs.
Policies
Use of the connector is governed by Martini’s Terms of Service and Privacy Policy. Generation jobs may also be subject to third-party model provider terms.
Troubleshooting
Browser-based MCP clients
Browser-based MCP clients connect from any origin; the connector authenticates every request with your bearer token, so no origin allowlisting is needed. If you see “Origin not allowed”, you are on a deployment that configured an explicit MCP_ALLOWED_ORIGINS allowlist — ask its administrator to add the client domain.
Uploads or downloads fail with a network error
If you’re calling Martini from an environment with restricted egress, asset storage may be blocked for both uploads (PUT) and downloads (GET). Run the check_storage_reachability tool first — it returns a probe URL and a remediation message you can follow.
Support
Connected AI agents can file feedback without leaving the session: call the submit_support_request tool to report a problem or suggestion directly to our support inbox. If you cannot connect, email us instead.
Questions, bug reports, and feature requests: support@martini.film. Partnership and integration inquiries: hello@martini.film.
More information is available on our support and security pages.