AI NPC gamesAI NPC game maker

AI NPC game maker for characters with memory

Capybara helps you build games where NPCs matter: villagers with routines, suspects with secrets, companions with context, and AI characters that react to game state instead of acting like generic chatbots.

Start buildingView pricing

Playable scene

AI NPC scene screenshot

Dialogue unlocked
100 24
Differentiator

AI NPCs should be grounded in the game world

The strongest AI NPC games do not start with an empty chat box. They start with a role, a place, a current objective, known facts, relationship values, inventory, clues, and limits on what the character will reveal.

Capybara’s planning guidance treats memory as a gameplay consequence. NPCs can remember gifts, promises, disrespect, secrets, debts, rumors, or town events, while the game still owns the facts that make quests and mysteries solvable.

Agent flow

Plan first, then build the NPC loop

The planning agent decides which NPCs matter, what they know, how they move, and what changes when the player talks, inspects, or chooses. After approval, the build chat wires the scene, dialogue UI, memory hooks, objective updates, and state changes.

First build

What Capybara can make first

Keep the first version small enough to finish and complete enough to play.

Playable slice

A playable top-down scene centered on two to four important NPCs with clear roles and movement patterns.

Dialogue connected to quests, clues, shops, relationships, access permissions, rumors, or objective updates.

Optional AI NPC memory, bounded free-text input, TTS narration, and code-owned ground-truth logs for mysteries.

Build flow

1

Give each NPC a job

Every character should be a suspect, witness, merchant, companion, quest giver, rival, guard, villager, or guide.

2

Connect talk to game state

Conversations should unlock clues, update quests, change prices, alter patrols, affect reputation, or reveal new choices.

3

Use memory carefully

Memory is strongest when it creates consequences: promises remembered, gossip spread, relationships changed, or secrets exposed.

How it works

Plan first, then build with the chat agent

Start with a plain-language idea. Capybara may ask a few focused questions when a decision changes the first map, NPCs, mechanics, or player goal.

Once the plan is ready, approve the first playable slice. The build chat uses that plan as the source of truth, then you keep improving the game by asking for clearer feedback, stronger NPCs, better objectives, new areas, or more polish.

Try it

Start with a prompt, then refine

Use one of these as a first message, or rewrite it in your own words.

Mystery NPCs

“Plan a top-down detective game with three AI NPC suspects, each with a motive, alibi, movement routine, bounded dialogue, a code-owned event log, clue inspection, accusation flow, and a result screen for correct or wrong accusation.”

Village memory

“Create a cozy village prototype where two AI villagers remember gifts and promises, update relationship state, give errands, and change dialogue after the player completes a delivery.”

Questions before you build

What makes Capybara good for AI NPC games?

Capybara plans AI NPCs around game state: role, place, objectives, inventory, evidence, relationships, memory, limits, and consequences. That keeps conversations playable instead of generic.

Can NPCs have schedules?

Yes. The first build can use patrols, post-work loops, shopkeeper stations, quest posts, or story-locked positions. Full daily schedules can be added later when they improve the loop.

Can AI NPCs be used in mysteries?

Yes, but mysteries should use a code-owned ground-truth event log so the case stays solvable. NPCs can lie or withhold information according to their role, while clues and objectives remain deterministic.

Ready to build this game idea?

Start with one scene and iterate from there.

Read blogSign up