Key Takeaways
- SillyTavern (github.com/SillyTavern/SillyTavern) is a free, open-source (AGPL 3.0) chat frontend with over 33,000 GitHub stars and 300+ contributors.
- It began in February 2023 as a fork of TavernAI, created by developer Cohee, with RossAscends joining shortly after as a core co-maintainer.
- It connects to local backends (KoboldCpp, text-generation-webui, Ollama, LM Studio) or cloud APIs (OpenAI, Claude, OpenRouter) β it generates no text itself.
- Core features include Character Cards (V1/V2/V3 spec), World Info/lorebooks, Group Chats with multiple characters, and Vector Storage for long-term memory.
- The project accepts no monetary donations for itself; it is maintained by volunteers, and applies no content filter of its own β that depends on whichever backend or model you connect.
π In One Sentence
SillyTavern is a free, open-source (AGPL 3.0) chat frontend built for AI character roleplay, with character cards, lorebooks, and group chats, that connects to a local backend or cloud API rather than generating text on its own.
π¬ In Plain Terms
SillyTavern is the interface, not the brain β it's the app you use to talk to a character, but a separate model (running locally or through a paid API) is what actually writes the character's replies.
Quick Facts: SillyTavern
- License: AGPL 3.0 (fully open source)
- Platforms: Web app, runs on Node.js; usable in any browser including mobile
- Website / GitHub: github.com/SillyTavern/SillyTavern β 33,000+ stars, 6,200+ forks, 300+ contributors
- Docs: docs.sillytavern.app
- Cost: Free; no company, no paid tier, project does not accept donations for itself
- Core function: Character-card and lorebook-based roleplay frontend that connects to a separately run model backend or API
What Is SillyTavern?
SillyTavern is a free, open-source chat frontend designed around persona-based roleplay rather than single-assistant chat. Instead of one generic "assistant" persona, it centers on Character Cards β structured files defining a character's personality, appearance, and speech patterns β and lets you talk to that character, or several at once in a Group Chat.
It has no built-in AI of its own. Every response is generated by a backend you connect separately: a local inference server like KoboldCpp, text-generation-webui, Ollama, or LM Studio, or a cloud API key from providers like OpenAI, Anthropic, or OpenRouter. SillyTavern's job is prompt construction, memory management, and presentation β not text generation.
What sets it apart from general-purpose chat frontends is depth of roleplay-specific tooling: World Info (lorebooks) that inject background facts only when relevant, Character Card V1/V2/V3 support with embedded lorebooks, regex-based prompt templating, and Vector Storage for retrieval-based long-term memory across long-running scenarios.
History: From TavernAI to SillyTavern
SillyTavern began as a fork of an earlier project called TavernAI, created by developer Cohee (GitHub handle "Cohee1207") on February 9, 2023, after development of the original TavernAI had slowed. RossAscends joined shortly afterward as a core co-maintainer, and Wolfsblvt became a regular contributor β together they have driven most of the project's direction since the fork.
Its first stable release after the fork is generally cited as v1.0 on April 9, 2023, though this specific date comes from a single secondary source and could not be independently cross-verified against a second primary record.
Early development relied partly on a companion "Extras" server β a separate Python microservice handling image captioning, text-to-speech, and vector-based features β which was marked obsolete and effectively discontinued around April 2024 once those capabilities were absorbed directly into the core application.
The Character Card specification evolved alongside the app: V2 added an embedded lorebook field (`character_book`) and alternate greetings, and the project now supports V3 as well, while remaining backward-compatible with V1 and V2 cards, which is why cards created years apart still generally load correctly.
As of this writing, the last release we could directly confirm is v1.18.0, published May 3, 2026, which added Cloudflare Workers AI as a supported backend alongside existing local and cloud options. Given SillyTavern's typical four-to-six-week release cadence, later 2026 releases are likely by the time you read this β check the releases page for the current version.
What Can You Actually Do With SillyTavern?
- Talk to structured AI characters, not just a generic assistant. Character Cards (V1/V2/V3) define personality, appearance, and speech style, and are widely shared and reused across the community.
- Keep long scenarios consistent with lorebooks. World Info entries inject background facts into the prompt only when a relevant keyword appears in the conversation, instead of bloating every single message with static backstory.
- Run scenes with more than one character. Group Chats let multiple character cards participate in the same conversation, each responding according to configurable turn-taking rules.
- Give long roleplay sessions long-term memory. Vector Storage retrieves relevant past messages by embedding similarity rather than relying only on however much recent history fits in the model's context window.
- Connect almost any model you want. Local backends (KoboldCpp, text-generation-webui, Ollama, LM Studio) and cloud APIs (OpenAI, Claude, OpenRouter, Cloudflare Workers AI, and others) are all supported through the same interface.
- Extend it with image and voice features. Multimodal image captioning and text-to-speech extensions add non-text capabilities on top of the core chat experience.
Who Should Use SillyTavern?
SillyTavern is built for people who want deep, granular control over AI-driven roleplay and interactive fiction β not for someone looking for a general-purpose assistant.
- β You want deep control over character personality, memory, and multi-character scenes β Character Cards, lorebooks, and Group Chats are exactly this.
- β You already run a local model (KoboldCpp, text-generation-webui, Ollama, LM Studio) or have a cloud API key β SillyTavern is one of the most widely supported frontends for connecting to any of them.
- β You want long-running scenarios that stay consistent over many sessions β World Info and Vector Storage are built specifically for this.
- β You are comfortable with a slightly technical setup (Node.js, backend configuration) β the payoff is far more configurability than turnkey roleplay apps offer.
- β You want a general-purpose assistant for productivity, coding, or research β use Open WebUI or Cherry Studio instead.
- If unsure: install SillyTavern, connect a free local backend like KoboldCpp with a small GGUF model, and load one of the widely shared example character cards before writing your own.
When Should You NOT Use SillyTavern?
Being a deep, backend-agnostic roleplay frontend is SillyTavern's whole purpose, and that focus is exactly why it is a poor fit outside that use case.
- If you want general productivity or coding assistance: SillyTavern's entire UI is oriented around character-based roleplay; a general assistant frontend will be simpler and more direct for that.
- If you want a single app with nothing to configure: you must separately install and connect a backend or pay for a cloud API key before SillyTavern can generate a single reply.
- If you need built-in content moderation: SillyTavern applies no content filter of its own; whatever the connected model or API allows is what will be generated β plan your backend choice accordingly.
- If you want guaranteed support or an SLA: it is a volunteer-maintained open-source project with no commercial support tier.
- If you want a turnkey experience with zero backend setup: consider a more turnkey local roleplay app like Backyard AI instead.
How Do You Install SillyTavern?
Setup has one prerequisite and a choice of two install paths, followed by connecting a backend:
- 1. Install Node.js (if not using Docker) β a current release for your OS.
- 2. Clone the release branch. `git clone https://github.com/SillyTavern/SillyTavern -b release`, then enter the created folder.
- 3. Start the server. Run `./start.sh` (Linux/macOS) or `Start.bat` (Windows) β dependencies install automatically on first run, and the app becomes reachable at `http://localhost:8000` by default.
- 4. Or use Docker instead. Run the `ghcr.io/sillytavern/sillytavern` image via `docker compose up`, mounting config, data, extensions, and plugins volumes as documented at docs.sillytavern.app/installation/docker.
- 5. Connect a backend. In the API Connections panel, add a local backend (KoboldCpp, text-generation-webui, Ollama, LM Studio) or a cloud API key (OpenAI, Claude, OpenRouter) β nothing generates until this step is done.
- 6. Load or create a character card, then start chatting.
- Full documentation is maintained at docs.sillytavern.app.
Pricing: Free, No Company Behind It
SillyTavern has one of the simplest pricing stories in this space: it is free, and there is no paid tier to eventually upsell you to.
- Free, forever, for everyone. SillyTavern is licensed under AGPL 3.0, maintained by volunteers, with no company, no subscription, and no paid version of the app itself.
- The project does not accept donations for itself. Its own materials state it has no servers to fund or hosting costs to cover, since it runs on your own machine.
- A co-maintainer accepts personal support separately. RossAscends runs a personal Patreon/Ko-fi with Cohee's blessing, but this supports the individual, not the project, and does not steer development direction or unlock features.
- Your real cost is whatever backend you connect. A local model through KoboldCpp or Ollama has no per-message cost beyond your own hardware; a cloud API key (OpenAI, Claude, OpenRouter) is billed by that provider, not by SillyTavern.
SillyTavern vs Agnaistic vs RisuAI vs Backyard AI: Which One?
All four target AI roleplay, but trade off configurability, hardware needs, and turnkey simplicity differently.
Tool | Link | Best for | Runs models itself | Character cards | Group chats | License |
|---|---|---|---|---|---|---|
| SillyTavern | github.com/SillyTavern/SillyTavern | Deepest configurability and backend choice for character roleplay | No (needs a backend/API) | V1/V2/V3 | Yes | AGPL 3.0 |
| Agnaistic (Agnai) | agnai.chat | Lighter-weight roleplay that runs well on older hardware | No (needs a backend/API) | Yes | Yes | Open source |
| RisuAI | risuai.net | Visual-novel-style presentation with sprites and branching dialogue | No (needs a backend/API) | Yes | Limited | Open source |
| Backyard AI | backyard.ai | Turnkey local roleplay app with less setup than SillyTavern | Yes (built in) | Yes | Limited | Free (closed source) |
For a general-purpose (non-roleplay) chat comparison, see Open WebUI vs SillyTavern. For a deeper three-way roleplay-frontend comparison, see SillyTavern vs Agnai vs RisuAI.
Common Mistakes With SillyTavern
- Installing SillyTavern and expecting it to respond without a connected backend. It generates nothing on its own β connect a local backend or a cloud API key first in the API Connections panel.
- Assuming SillyTavern filters or moderates content. It applies no content filter itself; behavior is entirely a function of the connected model or API and its own policies.
- Stuffing every detail into the character card instead of using World Info. Long static character descriptions bloat every prompt; use lorebook entries for facts that only matter in specific contexts.
- Confusing project donations with a co-maintainer's personal support page. The project itself does not solicit donations; an individual maintainer's personal Patreon is separate and does not fund "the project."
- Expecting it to work well for coding or research tasks. Its prompt construction and UI are tuned for character roleplay, not general assistant use cases.
Common Questions About SillyTavern
What is SillyTavern?
SillyTavern is a free, open-source chat frontend built around character cards, lorebooks, and multi-character group chats. It connects to a local model backend or a cloud API to generate responses; it does not run or host any model itself.
Is SillyTavern free?
Yes, completely. It is licensed under AGPL 3.0, has no company behind it, and the project itself does not accept monetary donations, stating it has no servers to fund. A co-maintainer runs a personal Patreon that supports them individually, separate from project funding.
Where did SillyTavern come from?
It began as a fork of TavernAI, created by a developer known as Cohee (GitHub handle Cohee1207) in February 2023, after TavernAI's original development had slowed. RossAscends joined shortly after as a core co-maintainer.
Does SillyTavern generate text by itself?
No. SillyTavern is a frontend only β it needs a connected backend such as KoboldCpp, text-generation-webui, Ollama, or LM Studio, or a paid cloud API key (OpenAI, Claude, OpenRouter, and others), to actually generate a response.
What are character cards and lorebooks in SillyTavern?
A character card defines a persona's personality, appearance, and speech style, following the Character Card V1/V2/V3 specifications. A lorebook (World Info) injects background facts into the conversation only when relevant keywords appear, keeping long-running scenarios consistent without bloating every prompt.
Can SillyTavern do group chats with multiple characters?
Yes. Group Chats let multiple character cards participate in the same conversation, taking turns or responding based on configurable rules.
Does SillyTavern have content moderation?
SillyTavern itself applies no content filter of its own β what a connected model will or won't generate depends entirely on that model's own training and any moderation built into the backend or cloud API you connect it to.
What is SillyTavern's Vector Storage feature?
Vector Storage retrieves relevant past messages using embeddings rather than only recent chat history, giving long-running roleplay sessions a form of long-term memory beyond the model's normal context window.
How is SillyTavern different from Agnaistic or RisuAI?
Agnaistic (Agnai) is a lighter-weight, simpler alternative that runs well on older hardware but offers less deep customization. RisuAI leans into a visual-novel-style presentation with sprites, backgrounds, and branching dialogue trees. SillyTavern is generally considered the deepest and most configurable of the three, at the cost of a steeper learning curve.
Is SillyTavern good for coding or productivity tasks?
No, that is not what it is built for. Its entire feature set β character cards, lorebooks, group chats β is oriented around persona-based roleplay and interactive fiction, not general assistant tasks. Use a tool like Open WebUI or Cherry Studio for that instead.
Sources
- SillyTavern source code, releases, and license β github.com/SillyTavern/SillyTavern
- SillyTavern official documentation β docs.sillytavern.app
- GitHub API repository data (creation date, license, star count)
- SillyTavern-Extras repository (marked obsolete) β github.com/SillyTavern/SillyTavern-Extras