Creative Patterns

Use cases

The full Reddit creative system: data in, evidence out, any model or pipeline.

Sync live or CSV, run a math-first pattern engine, turn winners into the next test or import sheet, and keep everything in Markdown you can paste into ChatGPT, script against over HTTP, or hand to an MCP host. This page walks the loop end to end.

The operating loop

Ingest

Pull Reddit truth into a workspace

No shadow spreadsheet—creatives, metrics, and line IDs in Campaigns/.

  • Reddit Ads API (OAuth) or CSV import; same tree either way
  • Brand & GTM notebooks sit beside every campaign

Intelligence

Score, cluster, and narrate the evidence

The engine does math first; AI only tells the story the numbers allow.

  • Winners, losers, lift vs baseline, confidence you can show a stakeholder
  • What to scale, pause, or isolate before you spend again

Act & ship

Hand off to humans, models, and systems

Runbooks and CSVs for the next test; files for any LLM; API & MCP in flight.

  • Markdown for ChatGPT, Claude, Cursor—paste or fetch via MCP
  • Read API today; more triggers and automation as the surface matures

How one campaign travels through

1. Sync2. Structure3. Patterns4. Decide5. Ship6. Integrate

Reddit data lands in a real workspace

Connect the official Reddit Ads API (authorized account) or drop a CSV—creatives, metrics, and line IDs normalize into Campaigns/ as Markdown and structured files next to Brand and Go To Market.

  • No second campaign builder: the workspace mirrors what you are already running.
  • Stable IDs and metrics so the pattern engine and runbooks use the same truth the UI does.
  • File-first teams stay on CSV; API teams get live pulls—same folder and notebook model.

Data path

  1. Connect Reddit or import CSV

    OAuth or file drop

  2. Normalize to workspace

    ads, metrics, copy, IDs

  3. Campaigns/ + sidecars

    MD, CSV, briefs

Find patterns you can stand behind

The pattern engine scores every slice: weighted CTR, lift vs baseline, confidence from sample size. You see winning and underperforming clusters, overlap hints, and mechanical cross-checks—not a black-box “AI said so.”

  • Optional OpenAI narratives on top: prose only, grounded in calculated evidence—no invented performance.
  • Same pass powers the Patterns tab and runbooks: what you read in the app is what the handoff sees.
  • Clarity for ‘what to test next’: hooks, audience splits, and creative structure surface as first-class objects.

Pattern pass

  1. Ingested metrics

    import + engine input

  2. Math + thresholds

    CTR, lift, confidence

  3. Cards + optional story

    facts, then narrative

Drive better Reddit decisions

Use evidence to optimize—not just report. Pattern cards and spend share tell you where to double down, what to pause or isolate, and which levers to pull (creative vs placement vs bid) before you re-spend.

  • Turn “winner / loser” slices into next actions: scale tests, holdouts, and quarter-style kill rules.
  • Narratives call out when clicks are expensive vs when CTR is misleadingly high—so you don’t optimize the wrong thing.
  • Runbooks and imports reuse the same campaign truth, so the next batch in Ads Manager lines up with what the engine proved.

Optimization loop

  1. Evidence in hand

    patterns + spend

  2. Decide the lever

    creative, audience, bid

  3. Launch next test

    runbook or import sheet

Notebooks and runbooks for *your* models

Everything lives in Markdown you can paste into ChatGPT, Claude, or Cursor—or pull via integrations. Runbooks bundle campaign truth, brand and GTM snippets, and an execution command so the model knows the job. Your notebooks are the context other AI tools never got from a CSV alone.

  • Timestamped runbooks under Runbooks/; strict copy, Meta carry, image/motion prompts, and landing tests when you need them.
  • The same pattern run the UI used is merged with notebook excerpts so external models are grounded.
  • Outputs are yours to diff, version, and ship—not trapped in our chat window.

Handoff

  1. Engine + brand context

    one composed brief

  2. Runbook / notebook files

    Markdown, CSV

  3. Any tool

    LLM, IDE, or agent host

Import-ready rows with traceability

Generate evidence-backed bulk rows: Creative Brief and Hypothesis per line, QA mapping to pattern IDs, and optional COPY_ONLY_TEST when you are text-first. Ship to Reddit Ads Manager with a paper trail, not a mystery CSV.

  • Batch rationale in plain language: what you are extending, what you are avoiding.
  • Image prompt cells optional; align writers with strict copy runbooks in the same workspace.
  • Same performance truth as the Patterns tab—traceable to the next launch.

Import flow

  1. Pattern + brief context

    engine output

  2. CSV + QA table

    rows + mapping

  3. Reddit Ads Manager

    import shape

API and MCP: wire Creative Patterns in

Treat the workspace as infrastructure. HTTP API (workspace API keys) exposes read paths for notebook files; MCP lets Cursor, Claude Desktop, and other agents pull the same Markdown and context without copy-paste. Write, triggers, and credit-aware jobs are on the roadmap for automation-heavy teams.

  • Create scoped keys in Integrations; use bearer auth the way your scripts and agents expect.
  • MCP config: point your host at the repo server with your workspace and key—live context in the agent thread.
  • V1 is read-biased; we are expanding the surface in lockstep with billing and credits on heavy actions.

Extensibility

  1. Workspace + key

    scope & revoke

  2. HTTP and MCP

    files, then more

  3. Your pipeline

    CI, agents, internal tools

Memory you own, everywhere you work

Git-friendly Markdown: brand, positioning, pattern intelligence excerpts, and runbooks compound over time. Diff, branch, and paste into any workflow—then layer API or MCP when you are ready to automate.

  • No rental dashboard: the artifact is a file tree you can export, back up, or open in the editor you already use.
  • Hooks, claims, and test history stay yours when the ad platform UI forgets last month.
  • Natural bridge to the API and MCP story: same content in the browser, in a script, and in an agent.

Portability

  1. Workspace files

    MD + CSV

  2. Version & share

    git, paste, link

  3. Stack of choice

    LLM, code, or API

Runbook templates (8)

Same pattern pipeline as Patterns, then composed Markdown or CSV. Credits debit on success; failed jobs refund.

  • Generate New Reddit Ads From Winning Patternsmarkdown
  • Strict Reddit Copy Only (From Winning Patterns)markdown
  • Turn Winning Hook Into Meta Ad Variantsmarkdown
  • Create Image Generation Prompts From Creative Learningsimage_prompts
  • Create Video / Motion Prompts From Creative Learningsvideo_prompts
  • Evidence-backed Reddit import sheetcsv_reddit
  • Site message tests (optional)markdown
  • Summarize Campaign Learnings for Future Testsmarkdown
Use cases — Reddit patterns, optimization, runbooks, API & MCP | Creative Patterns · Creative Patterns