Talk to it, don't command it
Mention Sophia, reply to her, or use /talk.
She keeps conversation continuity across reply chains and threads, so follow-ups just work. No command cheat-sheet required.
Your Discord server,
finally answerable.
Sophia is a conversational AI agent that searches your history, manages channels and roles, and cleans up messages — all from plain language, with admin approval on every action that changes your server.
Self-host free for non-commercial use. Hosted tier in early access.
PS
@Sophia what did Faymon say in #planning last week, and archive the old #q2-launch channel
Faymon posted three times in #planning on the 9th — the summary is that the launch slipped to July.
[jump ↗]
Approval required · destructive
Archive channel #q2-launch. This needs your approval and a confirmation.
Approve
Dismiss
Ask in plain language. She figures out the rest.
Sophia reads your server history to answer who said what, where, and when — and she can take action like creating a channel, assigning a role, or archiving old messages, without you memorizing a single slash-command syntax. Every write goes through an approval card. Nothing destructive happens without your explicit confirmation.
Deep server search
Sophia indexes your messages locally for fast lookups and automatically pulls live Discord history when the cache isn't enough. Answers come from real message evidence with jump links, not guesses.
33 tools, one agent
From retrieving messages and resolving members to creating channels, moving categories, managing roles, and editing or clearing messages — the model picks the right tools and composes them for your request. It's an agent, not a fixed script.
Approval gates that keep admins in control
Write actions need admin approval. Destructive actions need approval plus a confirmation dialog, and can never be auto-approved. Protected super-channels block destructive actions entirely. Role-based access splits full-control admins from scoped moderators, with rate limiting.
From question to done in three steps
No setup ceremony. Add her, ask, approve.
-
1
Connect
Add Sophia to your server. Self-host with your own Discord token and OpenRouter key, or join the hosted tier and skip setup entirely.
-
2
Ask
Mention her or use
/talk. She loads conversation context, searches indexed history, and refreshes from live Discord when needed. -
3
Approve & done
For anything that changes your server, Sophia shows an approval card. You approve (and confirm, if destructive). She executes and reports back.
Pricing
Self-host is a real product, not a crippled demo — read the code and run it yourself. The hosted tiers exist so you don't have to.
Self-Host
Open source, full control
€0
non-commercial
For technical admins who want it all. Bring your own Discord token and OpenRouter key, pay OpenRouter directly.
- ✓ Full source, all 33 tools
- ✓ All 17 model profiles
- ✓ Approval system & RBAC
- ✓ Local message indexing
- ✓ Community support on GitHub
Recommended
Hosted — Starter
Zero setup, we run it
€12
/ server / month
We run, update and meter Sophia for you. One Discord server.
- ✓ ~2M tokens / month included*
- ✓ Metered overage at cost + 15%
- ✓ Tuned default model profile
- ✓ Hosted usage dashboard
- ✓ Email support
Hosted — Pro
For large, busy servers
€39
/ server / month
Everything in Starter, with more headroom and the premium models.
- ✓ ~10M tokens / month included*
- ✓ All profiles incl. premium (GPT / Gemini Pro class)
- ✓ Priority queue & higher budgets
- ✓ Multi-admin access control
- ✓ Priority support & early tool access
* Token allotments and prices are early-access estimates, not committed SLAs — the dashboard shows live usage and overage is metered, never a surprise cutoff mid-conversation. Need multiple servers or SSO? Talk to us.
Frequently asked
Self-host vs hosted — what's the difference? +
Self-host is free for non-commercial use: you run Sophia on your own machine and pay OpenRouter directly for tokens. Hosted means we run and update Sophia for you, meter your token usage, and give you a dashboard — you pay a flat monthly fee per server with an included token budget.
Who can see my server data? +
Self-hosting, nothing leaves your machine except the LLM calls you make to OpenRouter. On the hosted tier, we process messages only to answer your requests and run indexing; we don't sell data. Message content sent to the model goes to your chosen LLM provider via OpenRouter under their terms.
Which AI models can I use? +
Sophia ships with 17 model profiles spanning Google, OpenAI, DeepSeek, Mistral, MiniMax, xAI, and free-tier models.
Swap models in /settings.
Self-hosters can use any of them; hosted Starter uses a tuned default, hosted Pro unlocks the premium profiles.
What does it cost to run? +
Sophia itself is free to self-host. The real cost is LLM tokens. Self-hosters pay OpenRouter directly and control spend by choosing cheaper or free model profiles and tuning tool-call and latency budgets. Hosted tiers bundle a monthly token budget with transparent, metered overage.
Will it delete things on its own? +
No. Write actions require admin approval. Destructive actions — clearing or deleting messages, deleting channels — require approval and an explicit confirmation, and can never be auto-approved. Protected channels block destructive actions outright.
Make your server answerable.
Run it yourself today, or get on the list for the hosted version and skip the setup entirely.