Product
D&D Session Recap and Prep
For the busy Dungeon Master running a home 5e / PF2e / OSR / Daggerheart game who has 15 minutes before the party rolls up and hasn't opened the campaign folder since last Tuesday. Paste your session notes (bullets, voice-memo transcript, or a rough recap), a one-line party roster, the campaign tone, the vibe you're going for next session, and (optionally) any thread you want emphasised. Get back a player-facing 'Previously on...' recap ready to read at the top of the table AND a Lazy-DM-shaped prep pack you can run from: a strong-start scene, exactly 10 secrets and clues, exactly 3 fantastic locations, 5-6 NPCs with a voice quirk and a hidden motive, unresolved threads pulled from your notes with callback-timing markers, and 2-3 lull-breaker hooks for when the table stalls. Hard no-invention guard: no NPCs, no locations, no plot beats that weren't in the notes textarea unless flagged as newly-created in the provenance map. Hard no-stat-block guard: never emits HP, AC, attack bonuses, damage dice, save DCs, spell slots, or CR values, combat threats are described in prose only. Hard no-player-agency guard: never scripts what the PCs do or feel; only sets up situations the DM presents. Deep tier adds per-NPC dialogue samples + physical tell, per-thread callback timing, a branching strong-start note, an entity_provenance_map so you can audit exactly what the model kept from your notes vs invented, and a full pass of the lazy-fantasy filler-word strip (whispered, ancient, forgotten, mysterious stranger, cloaked figure, tavern in a small village, the shadows danced, a chill ran down, hung heavy, an aura of, no ordinary).
About this tool
What does d&d session recap and prep do?
For the busy Dungeon Master running a home 5e / PF2e / OSR / Daggerheart game who has 15 minutes before the party rolls up and hasn't opened the campaign folder since last Tuesday. Paste your session notes (bullets, voice-memo transcript, or a rough recap), a one-line party roster, the campaign tone, the vibe you're going for next session, and (optionally) any thread you want emphasised. Get back a player-facing 'Previously on...' recap ready to read at the top of the table AND a Lazy-DM-shaped prep pack you can run from: a strong-start scene, exactly 10 secrets and clues, exactly 3 fantastic locations, 5-6 NPCs with a voice quirk and a hidden motive, unresolved threads pulled from your notes with callback-timing markers, and 2-3 lull-breaker hooks for when the table stalls. Hard no-invention guard: no NPCs, no locations, no plot beats that weren't in the notes textarea unless flagged as newly-created in the provenance map. Hard no-stat-block guard: never emits HP, AC, attack bonuses, damage dice, save DCs, spell slots, or CR values, combat threats are described in prose only. Hard no-player-agency guard: never scripts what the PCs do or feel; only sets up situations the DM presents. Deep tier adds per-NPC dialogue samples + physical tell, per-thread callback timing, a branching strong-start note, an entity_provenance_map so you can audit exactly what the model kept from your notes vs invented, and a full pass of the lazy-fantasy filler-word strip (whispered, ancient, forgotten, mysterious stranger, cloaked figure, tavern in a small village, the shadows danced, a chill ran down, hung heavy, an aura of, no ordinary).
How does D&D Session Recap and Prep work?
Paste your last session's notes (bullet events, voice-memo transcript, or a rough recap), one line per PC with name and class and level, the campaign tone (grimdark, high-fantasy, weird-west, horror, comedy-heist, or noblebright), the next-session vibe (combat-heavy, investigation, social-court, travel-hex, dungeon-crawl, or boss-fight), when the session is (tonight, tomorrow, this-week), and optionally a DM note. One structured pass returns a player-facing 'Previously on...' recap ready to read at the top of the table plus a Lazy-DM prep pack you can run the session from: a strong-start scene, exactly 10 secrets and clues spread across backstory, foreshadowing, immediate, and PC-connected kinds, 3 fantastic locations with a sensory hook and an interactable element each, 5 to 6 NPCs with a voice quirk and a hidden motive, 3 to 5 unresolved threads pulled straight from your notes, 2 to 3 lull-breaker hooks, and 3 to 4 combat challenges described in pure prose. Every element is defensible line by line against the notes you pasted.
How is this different from asking ChatGPT to prep my session?
A freeform ChatGPT prompt happily invents NPCs, locations, and plot beats that were never in your notes, drifts into scripting what the players do or feel, and spits out HP, AC, CR, save DCs, and damage dice with wrong math you then have to fix at the table. D&D Session Recap and Prep runs a single structured pass with three hard guards: a no-invention guard that refuses to introduce named NPCs, locations, or plot beats that were not in your notes (except explicitly flagged newly-created entries in the provenance map on the deep tier), a no-stat-block guard that refuses to emit HP, AC, attack bonuses, damage dice, save DCs, spell slots, or CR values anywhere (combat challenges are described in prose only), and a no-player-agency guard that refuses to script what the party does, feels, decides, or says. Every guard rewrite is receipted in the safety_notes panel so you can verify the pass ran before you sit down at the table.
Will it invent NPCs, locations, or plot beats that were not in my notes?
No. Every named NPC, every location, every unresolved thread, and every combat entity you see in the output is either drawn directly from the notes you pasted or flagged as newly-created in the entity provenance map on the deep tier. The provenance map shows one entry per named entity with a kind of from_notes or newly_created and, for from_notes entries, the exact verbatim snippet from your notes it was drawn from. If the recap would need to fabricate a plot beat to feel complete, the tool keeps the recap short and honest instead of padding it with invented flavour, and receipts the thin-notes case in safety_notes.
Will it produce stat blocks with invented HP and AC and CR?
No. There is a hard rule that refuses to emit HP, AC, attack bonuses, damage dice, save DCs, spell slots, spell save DCs, CR values, PF2e level values, initiative bonuses, or any numeric mechanical stat in any field. Combat challenges are described in prose only: what the creature opens with, how it fights, when it flees, what the party can exploit. Your monster manual owns the math; the tool supplies the fiction and the tactical read. Every stat-block phrasing the model catches itself typing gets rewritten to prose and logged in safety_notes.
What is the Lazy-DM prep formula and why does the tool use it?
The Lazy Dungeon Master by Sly Flourish is a widely used prep framework whose canonical five-minute shape is a strong start, exactly 10 secrets and clues, 3 fantastic locations, and a handful of NPCs. It is minimal, table-ready, and improvisation-friendly. This tool composes exactly that shape and adds three complementary pieces the busy DM also needs: unresolved threads pulled straight from your notes for continuity, lull-breaker hooks for when the table energy dies, and combat challenges the DM can drop in as pure prose. You get one artefact you can run the session from, calibrated to the tone and vibe you asked for.
Does it work for Pathfinder 2e, OSR, and Daggerheart, or only D&D 5e?
Yes to all of them. The tool never emits system-specific stat blocks or mechanical numbers, so the output is portable across 5e, Pathfinder 2e, OSR retroclones, Daggerheart, Shadow of the Weird Wizard, or any other tabletop fantasy system with a Dungeon Master. Party composition is one line per PC (name plus class plus level) so the tool tunes NPC vibes and combat challenges to a level of abstraction your system can translate. The DM handles the mechanical layer; the tool handles the fiction layer.
What if my last session notes are just three bullets or a voice memo transcript?
Thin notes are fine. The tool keeps the recap short and honest rather than padding it with invented plot beats, and receipts the thin-notes case in safety_notes so you can see the pass ran. Fantastic locations, NPCs, and combat challenges rounding out the prep pack are permitted to be newly-created (the deep tier flags each one in the entity provenance map so you can see exactly what came from your notes versus what was invented for this session). The recap in particular stays 100 percent anchored to what actually happened at the table; the prep pack can lean creative to fill the gaps.
Which tier should I pick, Quick or Deep?
Pick Quick (15 produde tokens on Gemini Flash Lite) when the party is 15 minutes out and you just need something table-ready: full recap, strong start, all 10 secrets, 3 locations, 5 or 6 NPCs, unresolved threads, lull hooks, combat challenges, no-invention guard, no-stat-block guard, no-player-agency guard, and a top 8 lazy-fantasy filler strip. Pick Deep (30 produde tokens on Gemini Flash) for the session you want your players talking about after: everything in Quick plus per-NPC dialogue samples calibrated to your campaign tone, per-NPC physical tells you can pantomime at the table, per-thread callback timing markers (drop-early, drop-mid-session, drop-only-if-lulls), a branching strong-start note calibrated to which way the party is likely to go, an entity provenance map so you can audit exactly what came from your notes versus what was invented, and the full 2026 lazy-fantasy filler lexicon stripped with named-swap receipts.
Is it free, and do I need a subscription?
There is no subscription. You spend tokens per run: 15 produde tokens for the Quick tier and 30 for the Deep tier. Tokens are a prepaid platform credit you buy once and spend across any produde tool, so you pay per session prep instead of a monthly TTRPG-tools SaaS fee that keeps charging you when your campaign is between arcs. If a run fails, the full token cost is refunded to your wallet automatically.
What does it cost?
From 15 tokens. Pick a tier on the form above. Produde tokens are a prepaid platform credit; buy them once and spend on any tool.
Do I pay if a run doesn't complete?
No. You only pay for results. If a run can't finish, its full token cost stays in your wallet automatically.
Is my data stored?
Your input and the run output are saved to your account so you can browse history. They are not sold, shared with advertisers, or used to train shared models. See privacy for details.