Skip to main content
Parlay’s playbook drafting service synthesizes one or more source materials (call notes, transcripts, methodology docs, battle cards) into a structured markdown playbook using Gemini 2.5 Flash. The draft is reviewable; you promote it to “active” with a single call.

​
The flow

  1. Upload source materials — N free-form text blobs, each with a title
  2. Generate a draft — async job synthesizes them into one markdown playbook
  3. Review — read the draft, request more if you want
  4. Publish — promote to the org’s active playbook (auto-applied to every future analysis)

​
When to use this

Good fits

  • Onboarding a new partner — turn their existing collateral into a Parlay-formatted playbook
  • Refreshing an outdated playbook — feed in recent top-performer call notes
  • Generating role-specific playbooks (SDR vs AE vs CSM) from segmented inputs

When to skip

  • You already have a polished playbook in markdown — use create_playbook instead
  • You want to fine-tune the playbook prose word-by-word — generate, then update_playbook to edit
  • The org has zero source material — the model needs something to synthesize from

​
Step 1 — Upload source materials

# First source
curl -X POST \
  https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbook-sources \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "title": "Top performer highlights",
    "content": "When customers push back on pricing, our top reps bridge to value by asking: what would solving this be worth in the next 12 months? They always confirm decision-making authority before scheduling a demo..."
  }'

# Second source
curl -X POST \
  https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbook-sources \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "title": "Objection handling cheat sheet",
    "content": "Common objections and proven responses: 1) It costs too much → reframe to monthly cost ÷ ROI. 2) We already have a solution → dig into what is not working with the current setup..."
  }'
Each source is up to 500,000 characters. You can upload 1 or 50 — the draft job accepts up to 50 source ids.

​
Step 2 — Kick off the draft

curl -X POST \
  https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbook-drafts \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "source_material_ids": ["<src1_id>", "<src2_id>"]
  }'
Response: 
{
  "draft_id": "b7ed4724-4204-4833-abe3-792d451371f1",
  "status": "queued",
  "created_at": "2026-04-24T18:53:17.867123+00:00"
}

​
Step 3 — Poll until complete

curl https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbook-drafts/<draft_id> \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY"
Typical latency: 15–30 seconds. When status === "completed": 
{
  "id": "b7ed4724-...",
  "status": "completed",
  "generated_content": "# Sales Playbook\n\n## Discovery Questions\n\n*   **Current Solution Gaps:** When a prospect indicates they already have a solution, dig into what specifically isn't working...\n\n## Budget Authority\n\nBefore scheduling a demo, confirm decision-making authority...\n\n## Objection Handling\n\n### Pricing pushback\nReframe price as ROI: \"What would solving this be worth in the next 12 months?\"\n...",
  "gemini_model": "gemini-2.5-flash",
  "gemini_input_tokens": 412,
  "gemini_output_tokens": 1853,
  "cost_usd_cents": 1
}
The generated_content is the full markdown playbook, ready to publish.

​
Step 4 — Review and publish

If the draft looks good: 
curl -X POST \
  https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbook-drafts/<draft_id>/publish \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY" \
  -H "Idempotency-Key: $(uuidgen)"
This:
  • Inserts the generated_content into the playbooks table as version 1, is_active=true
  • Marks the draft as published, sets published_playbook_id
  • Fires a playbook.published webhook
  • Returns { playbook_id, draft_id, published: true }
From here on, every analysis under demo-acme automatically uses this playbook in the analysis prompt.

​
Don’t like the draft?

You have a few options:
  • Generate a new draft with different sources or a different combination
  • Edit the draft externally, then create_playbook directly with your edited content (skipping publish_playbook_draft)
  • Publish anyway, then update_playbook to edit the content. This bumps the playbook version and snapshots history.
Drafts can’t be republished once they’ve been published once.

​
Doing this from MCP

For org "demo-acme", upload these two source materials [paste content], then generate a playbook draft from them. Show me the first 1,000 characters of the generated content. If it looks good, publish it.
The model calls upload_playbook_source (×2), generate_playbook_draft (auto-polls), shows you the preview, and (if you confirm) calls publish_playbook_draft.

​
Cost + latency

OperationCostLatency
upload_playbook_source$0under 1s
generate_playbook_draft~0.01–0.01–0.0515–30s typical, up to 3 min worst case
publish_playbook_draft$0under 1s
Cost scales with input + output tokens. Two short sources (a few hundred chars each) → ~0.01.Twentylongsources(50kcharseach)→ 0.01. Twenty long sources (50k chars each) → ~0.05.

​
What happens to the previous active playbook?

publish_playbook_draft does not automatically deactivate older playbooks. If you want only the new one active, follow up with: 
curl -X PATCH \
  https://parlay-api-dev-o7nogixtqq-uc.a.run.app/v1/orgs/demo-acme/playbooks/<old_playbook_id> \
  -H "Authorization: Bearer pk_sandbox_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{ "is_active": false }'
In a future release we’ll add an option on publish_playbook_draft to deactivate other active playbooks atomically.