The relationship layer
A campaign knowledge graph stores your world as typed entities, NPCs, factions, locations, quests, and items, connected by real relationships instead of prose. Grimoire builds the graph as you play: record a relationship once and it becomes a traversable edge you can view as a political web, a timeline, or a map.
Knowledge graphs are on the free tier. No credit card, no time limit.
The definition
A campaign knowledge graph is a structured map of everything in your campaign and how it connects. Each NPC, faction, location, quest, and item is a typed node; every relationship between them is an explicit edge. Instead of re-reading notes to remember who owes whom, you query the graph and get an answer.
Knowledge graphs are how large organizations map entities and relationships across sprawling data. A six-month D&D campaign has exactly that shape: NPCs who owe each other things, factions with overlapping interests, locations that change hands, items with ownership chains, and plot threads braiding through all of it. A document flattens that network into prose. A knowledge graph keeps it a network.
The important part is what the nodes and edges are made of. In Grimoire, a node is not a page title, it is a typed entity with structured fields: motivations, secrets, faction ties, a visibility tier. An edge is not a hyperlink, it is a named relationship. That is what makes a TTRPG knowledge graph queryable rather than just viewable, and it is the foundation everything below is built on.
The difference that matters
Wiki pages store prose and link when one page mentions another. Typed entities store structured fields, and their relationships carry meaning: member of, sworn enemy, located in. A backlink tells you two pages touch; a typed edge tells you why. That difference is what makes a campaign graph queryable rather than just browsable.
Grimoire ships fourteen typed entity schemas, from NPCs and factions to session recaps and world rules, so the structure exists before your first entry does. Fill in the duke's motivations and faction ties and the graph already knows how he connects to the rebellion, because the relationship is data, not a sentence somewhere in a document. See the full feature inventory for what each entity type carries.
This is the philosophical split between a campaign database and a campaign wiki: structured data you query versus articles you publish. Wiki-first tools are genuinely good at the article side. If you are weighing the two approaches, the LegendKeeper comparison and the World Anvil comparison walk the tradeoff honestly.
One graph, many lenses
Grimoire renders one campaign database through three built-in graph projections: a political web of NPCs, factions, and player characters; a campaign timeline anchored to your session recaps; and a geographic hierarchy of every place in your world. Each projection is generated from the same entities, so nothing is drawn or maintained by hand.
Who stands where, right now
Every NPC, faction, and player character, plus the relationships between them. Faction membership is one kind of edge; the interesting ones are specific: "related (mother)," "despises," "planning to betray." The political web is the live status of allegiance in your campaign, not a seating chart you drew once and forgot.
What happened, in order, with receipts
A session-by-session record where timeline nodes connect to real entities. Link the locations visited during Session 12 to Session 12 and the timeline shows the party's path, character development, and territory changes across months of play. Key events can be marked canon: facts the campaign can no longer misplace.
What sits inside what
Your locations in a hierarchy, plane to world to continent down to a single shop, with spatial relationships between siblings. NPCs and items sit inside that hierarchy, so "where is the Ashen Blade right now" has a structured answer. Upload images (a region, a city plan, a dungeon floor) and pin entities on them.
A fourth lens is yours to define: custom graphs let you assemble a projection around any axis you care about. A heist crew, a pantheon, a trade network between three port cities. Underneath them all, World Foundations holds the deep canon: pantheons, laws of magic, the bones of the setting.
At the table
GMs use a campaign knowledge graph for three jobs: session prep that starts from connections instead of a blank page, consistency checks that catch contradictions before players do, and controlled reveals, since every node carries a visibility tier that decides what the player portal shows. All three come from recording relationships once.
Open the duke and see every faction he belongs to, every location he holds, every quest that touches him, and every NPC tied to those factions. Three sessions in, the graph starts showing you connections you forgot you made. Prep becomes pulling threads, not staring at a blank page.
Canon lives in structured fields and typed edges, so contradictions surface instead of hiding in prose. When the new betrayal does not square with a recorded alliance, the graph shows both edges side by side. Your world stays coherent because it is data you query, not documents you re-read at midnight.
Every node carries a visibility tier: common knowledge, player knowledge, or GM secret. The tiers filter every projection, so the player portal shows your table the revealed slice of the same graph you run the campaign from. Flip one tier and the traitor's true allegiance appears on your players' map.
The free tier includes knowledge graphs, unlimited entities, and the player portal.
One full working campaign, forever. No credit card. No countdown.
Start your campaign. Free, no card. →Optional
Grimoire runs fully without a single AI feature. If you already use Claude, ChatGPT, or another MCP-compatible client for prep, Grimoire MCP serves this same knowledge graph as structured data your AI queries live: it surveys the map of your world, pulls one projection in full, then deep-dives a single NPC, instead of guessing from pasted notes.
You bring your own client, so there is no token markup and no model lock-in, and the assistant answers from your canon because the graph is the source it reads. Turn it on when you want it. Ignore it forever if you do not.
See how Grimoire MCP works →FAQ
A campaign knowledge graph is a structured map of a TTRPG campaign: every NPC, faction, location, quest, and item is a typed node, and every relationship between them is an explicit, labeled edge. Because the graph is built from structured data, you can query it ("every NPC tied to the Vale faction") instead of re-reading prose notes to reconstruct who connects to whom.
Backlinks record that two pages mention each other. Knowledge-graph edges record what the connection actually is: member of, sworn enemy, located in, opened in Session 12. Grimoire uses those typed edges to render political webs, timelines, and geographic hierarchies from the same campaign data, which a mention-based link graph cannot do.
No. Grimoire's knowledge graphs are projections of your campaign database, not diagrams you arrange by hand. Record a relationship on an entity once and the edge appears in every projection it belongs to. Custom graphs let you hand-assemble a view when you want one, but the built-in political, timeline, and geography projections maintain themselves.
Yes, through the player portal, filtered by visibility tiers. Every entity carries one of three tiers: common knowledge, player knowledge, or GM secret. Players see the slice of the graph you have revealed and nothing else. Flip one tier when the party uncovers a secret and their view of the world updates where they stand.
No. D&D is the most common system on Grimoire, but the graph is system-agnostic and genre-aware. Pick Fantasy, Sci-Fi, Horror, Post-Apocalyptic, Modern, Superhero, or Historical and the entity vocabulary reshapes to fit; the political, timeline, and geography projections work the same way in every genre.
Record the relationships as you play. The connections you forgot you made come back on their own.
Free tier, no credit card. Knowledge graphs included from the first entity.