Connect your AI agent to BeatBandit's story engine
Develop stories, draft screenplays, and generate shot lists by asking ChatGPT, Claude, Cursor, or any other MCP-aware AI assistant. Your assistant stops just talking about your script and starts actually changing it — in your real BeatBandit project, with the same auth and credits as the website.
https://beatbandit.ai/api/mcpUse this URL in Claude Code, Cursor, VS Code, or any Streamable HTTP MCP client.
Why connect your AI assistant?
The Model Context Protocol (MCP) is how modern AI assistants reach beyond their chat window and call into real apps. Without it, asking an assistant to “help with my screenplay” gets you suggestions inside the chat. With it, the assistant can read your real project, run BeatBandit's wizards as backend jobs, propose surgical edits to specific scenes, and even render shot frames — without you ever opening beatbandit.ai.
It's not an either/or with the website. Use both. Your assistant brings general intelligence and the conversational interface; BeatBandit brings deep, domain-specific knowledge of how stories actually work — templates, beat structures, scene ordering, screenplay grammar, room review, treatment generation. The combination is stronger than either side alone, and you can switch between the web UI and your agent mid-project with no state to sync.
Your screenplay, story canon, scene graph, credits, and project history all stay where they are. The assistant just gets a stable, secure way to talk to them.
A few things to ask your agent
Once you've installed the server, try opening Cursor, Claude Code, or ChatGPT and saying any of these out loud:
Bootstrap a story from a one-line idea
Hand the agent a logline. BeatBandit's story bootstrap workflow generates a beat sheet plus a scene outline, scoped to whatever template you pick (movie, one-scene short, episode, series).
“Develop a movie from this idea: 'A retired chef discovers their old kitchen is a portal to historical food crimes.'”
Get notes from a virtual writers’ room
Run a Room Review and the agent returns structured notes from six writer-room personas: showrunner, broadcaster, network exec, head writer, audience lens, market lens.
“Run a room review on my pilot and surface the top three rewrite priorities I should fix before the next pass.”
Generate first drafts inside your assistant
The Screenplay Writer runs as a backend job. Your assistant kicks it off, polls for progress, then hands back the document so you can keep iterating.
“Generate the screenplay for my one-scene short and show me the draft.”
Iterate scenes without leaving the chat
Propose surgical edits to specific scenes, beats, dialogue, or descriptions. Nothing is committed until you approve, so you stay in control of the canon.
“Tighten the dialogue in scene 5 by about 30 percent without changing the story beats.”
Storyboard from prose
Generate a shot list for any scene, then render reference frames and stitch them into preview video — all chained in a single conversation.
“Build a shot list for scene 12 and render the first six frames in a film-noir style.”
Use the assistant you already use
Cursor, Claude Code, ChatGPT, Windsurf, Cline, VS Code — the same MCP server works in all of them, against the same project, with the same auth and credits as beatbandit.ai.
“Try the same prompt from a different client. The project state stays in sync.”
BeatBandit answers with structured progress, resource links, and proposals you can inspect — not just chat. Workflows that take real time (developing a story, writing a screenplay, generating a shot list) run as backend jobs, so you can keep chatting while they finish.
Plug it into your workflow
The MCP server is a stable, OAuth-protected interface to the same backend that powers beatbandit.ai. That makes it useful at three very different scales.
For individual writers
Bring whatever assistant you already use. BeatBandit becomes the story-domain layer your agent can actually call — templates, beat structures, scene logic, screenplay grammar. Use the website for visual inspection of beats, scene cards, and screenplay; use your agent for generation, revision, and one-off edits. Same project, same canon, no exporting or syncing in between.
For studios, writers' rooms, and small teams
Multiple writers, multiple clients, one shared project. One person can run a Room Review from Claude Code while another tightens dialogue from Cursor while a third reviews the result in the web UI. Auth, RLS, and credits are the same across every client and the website, so collaboration stays auditable without any custom plumbing.
For companies and integrators building content pipelines
If you're building agentic video, marketing-content, narrative-game, or interactive-storytelling systems, treat BeatBandit's MCP as the story development expert behind your stack. Your orchestration layer can chain:
- bootstrap a story from a brief or marketing prompt
- generate a screenplay or episodic structure as a backend job
- produce a shot list, then render reference frames and preview video
- hand the result back into your downstream production system
Everything is rate-limited, credit-metered, and idempotency-aware out of the box. Build once, plug into any agent host, scale alongside your own pipeline.
Quick install
The BeatBandit MCP endpoint lives at https://beatbandit.ai/api/mcp. Authentication is browser-based OAuth on first use — you sign in at beatbandit.ai once, and the client stores a token for subsequent calls.
Cursor · Windsurf · VS Code
Add to your workspace .cursor/mcp.json:
{
"mcpServers": {
"beatbandit": {
"url": "https://beatbandit.ai/api/mcp"
}
}
}Claude Code
Register the server via the Claude CLI:
claude mcp add --transport http beatbandit https://beatbandit.ai/api/mcpPrompt pipelines
5 public prompts. These are BeatBandit-authored orchestration templates that inspect live project state and steer the client toward the right tools and resources.
Idea To Screenplay Pipeline
idea_to_screenplay_pipeline
Route a BeatBandit project from raw idea through outline stages into screenplay generation, branching by template mode and live readiness instead of only raw card counts.
Review To Revisions Pipeline
review_to_revisions_pipeline
Start with BeatBandit-led diagnosis first, then route the resulting review findings into the right revision path.
Revision Strategy Pipeline
revision_strategy_pipeline
Route a known revision request to the right screenplay-first or story-first BeatBandit change surface, route Scene List Revision merge survivors through reviewable scene authoring, and route remaining structural-sync queues through Sync Screenplay (`story_review_run` continuity_audit) before more prose edits.
Screenplay To Novel Pipeline
screenplay_to_novel_pipeline
Route an authorized Movie, Genre Movie, or existing Novel project through the canonical source → Blueprint → Chapters → Voice → manuscript → audit → export lifecycle using live readiness.
Screenplay To Video Pipeline
screenplay_to_video_pipeline
Guide BeatBandit through the canonical screenplay-to-video pipeline for one scene or episode scope, keeping identity/art-direction references distinct from reviewed realized-continuity checkpoints.
Tool families
33 public tools, grouped by the workflow stage they target. Each tool has its own reference page with input/output schemas, examples, and when-to-use guidance.
Core Project
Discover, create, and read BeatBandit projects and templates.
Story Graph
Read and mutate the story-graph primitives: cards, scenes, and the graph itself.
Screenplay
Read, propose, and apply changes to structured screenplay documents.
Story Workflows
Run backend-owned parent workflows for development, revision, review, and screenplay generation.
Story Collaboration
Collaborate with BeatBandit through conversational guidance and persona-driven passes.
Shot / Media / Export
Generate shot lists, images, videos, and export artifacts.
Asset Query
Filter generated media and shot-element libraries without leaving the MCP surface.
Jobs
Poll async workflow progress through the canonical job surface.
What is the MCP server?
The Model Context Protocol lets AI agents call into external services through a stable contract of tools and resources. BeatBandit's MCP server is an adapter layer over the same backend that powers beatbandit.ai: your agent can discover projects, mutate the story graph, propose and apply screenplay edits, run wizards as backend-owned parent jobs, and generate shots/images/video — all against the same auth, credits, and RLS rules as the product UI. Public prompts sit above that tool surface as curated orchestration entrypoints; they do not replace the underlying tools.
Prefer high-level workflows (e.g. story_development_run ) when the intent is concept-to-outline. Fall back to low-level card/scene mutations for deliberate manual editing.