OG Character AI responses with proper memory and minimal filters.
Benchmarks
meisan runs on c137's memory engine — the numbers are theirs.
Same engine underneath. Pick the mode that fits the moment — meisan keeps everything in scope, no cross-contamination.
One-on-one with a single character — your partner, crush, or comfort character. Build them once and meisan remembers everything, so the relationship actually grows. No resets.
A scene you're part of — a whole cast or group chat, or you dropped in as your own character. Each roleplay keeps its own memory, so every world stays its own.
A story told to you in third person, like a novel that unfolds as you go. The world and cast stay consistent — every detail remembered as the story builds.
Not seeing them? Make your own — start from a blank slate, or just describe them in a sentence and meisan builds the rest.
Create a character →Every feature, every plan. Upgrade when you need more.
Free to use right now while I set up an adult-friendly payment processor. These plans go live once that's done — no charges until then.
Everything you need
First month, then £10/month
For regular use
Everything you need to know about meisan.
Three things: real persistent memory that actually keeps your character and story building over time, the necessary content filters kept minimal so they don't get in the way of the experience, and optional self-set usage controls so you can pace yourself if you want to — fully off by default.
Yes. Every chat has a memory panel — open it any time to see what's been learned, lock canonical facts so the model can't drift, edit characters and world details, or wipe a memory line you don't want kept.
Yes. Start a new chat, pick Import, choose the mode (chat / roleplay / story), copy the prompt into your Character.AI, Janitor, Chai, or SillyTavern chat, and paste the response back. meisan brings the character, scenario, world and cast across. Your conversation history stays in the source app — what you get here is a fresh chat with the same character and real memory underneath.
Yes. Every new account gets 5,000 free credits — around 500 messages, no card required. Credits scale by model size; cheaper models use less, larger ones use more. Subscribe when you run out.
Included models cover small, mid, and large — Rocinante 12B, UnslopNemo 12B, Cydonia 24B, Skyfall 36B, Euryale 70B, plus GLM 4.7 Flash, DeepSeek V4 Flash, Grok 4.3, and Gemini 3 Flash. Bring your own OpenRouter key to use any other model — Claude, GPT-5, Llama, anything OpenRouter supports.
Yes. Conversations stay in your account — never sold, never shared. Replies are generated through third-party model providers, and meisan only uses providers whose API policies bar training on your messages (routing is locked to non-training providers). meisan may use de-identified data to improve the product, including its models — switch this off anytime in Settings → Data & Privacy. Content blocked by the safety system is kept briefly for review. Everything is deletable and exportable from settings.