The Agnai Character Format Explained (and How to Migrate)

Agnai is a flexible, multi-user frontend with a character schema that differs from the Tavern norm in a few important ways. Understanding those differences is the key to migrating Agnai characters cleanly.

Agnai stores characters as JSON, but with a schema that diverges from Tavern cards in three notable ways: the persona is an object rather than a plain string, the greeting is stored under greeting, and example dialogue lives under sampleChat. Those differences are exactly where naive conversions go wrong, so knowing them makes migration reliable.

How Agnai structures a character

An Agnai character JSON includes familiar top-level fields like name, description, and scenario. The divergence is in three places. First, persona is not a string — it is an object with a kind (such as "attributes" or "text") and an attributes map holding lists of traits. Second, the opening message is greeting, not first_mes. Third, example dialogue is sampleChat, not mes_example.

This design reflects Agnai's flexibility — the attribute-based persona supports different persona "formats" the app can render in various ways. But it means a converter that only looks for Tavern key names will miss the greeting and examples, and will treat the persona object as opaque.

Why a careless conversion produces a half-empty card

If you export an Agnai character and run it through a tool that expects Tavern field names, you can end up with a card that has the right name and scenario but a blank greeting and no example dialogue — because those tools looked for first_mes and mes_example and found greeting and sampleChat instead. The persona object may also be dumped in as raw structured data rather than readable text. The character imports, but hollowed out.

Converting Agnai characters the right way

A correct migration maps Agnai's fields onto the standard model: greeting becomes the greeting, sampleChat becomes the example dialogue, and the object persona is flattened into readable personality text by joining its attribute lists. Once that mapping is done, the character can be written into any target format — a V2 PNG for SillyTavern, JSON for Voxta, or a V3 CharX. When you convert, verify the greeting and examples specifically, since those are the fields most likely to be affected by schema differences.

Migrating to Agnai

Going the other direction is more forgiving. When you convert a Tavern card to Agnai, the standard fields map onto Agnai's structure, and the persona can be provided as text that Agnai accepts. Agnai's willingness to read a plain persona makes importing into it straightforward compared with exporting out of it.

Agnai's strengths worth keeping in mind

Agnai is notable for being genuinely multi-user and multi-backend: it supports shared chats and a range of AI services, and it can run self-hosted. That flexibility is why some creators keep characters there. The trade-off is the schema divergence covered above — worth the small extra care at conversion time.

Understanding Agnai's persona formats

Agnai historically supported several persona "formats," and the object-based persona reflects that. An "attributes" persona holds named lists — for example a personality list and a speech list — that Agnai can render into the prompt in different styles. A "text" persona is closer to the Tavern approach: a single descriptive block. Knowing which kind your character uses tells you how a converter should flatten it: attribute lists get joined into readable sentences, while a text persona passes through almost directly.

This is also why round-tripping an Agnai character through a Tavern-only tool and back can subtly change it: the structured attributes get flattened to text on the way out, and there is no clean way to reconstruct the original lists on the way back. If you plan to keep a character in Agnai long-term, keep the original Agnai JSON as your master copy and treat conversions as one-way exports for use in other apps.

Practical migration checklist

To move an Agnai character out cleanly: export the character JSON; convert it with a tool that understands greeting, sampleChat, and object personas; open the result in your destination app; and confirm four things — name, personality text (not raw attributes), greeting, and example dialogue. If any field is blank, the conversion likely missed an Agnai-specific key, and re-running through a normalizer that handles the Agnai schema will fix it. Our Universal Converter and the conversion guide cover the full process.

Why does my Agnai character import with a blank greeting?

Agnai stores the greeting under the key "greeting" and examples under "sampleChat," not the Tavern names first_mes and mes_example. A converter that only checks Tavern keys misses them. Use a tool that understands the Agnai schema.

How is the Agnai persona stored?

As an object with a kind and an attributes map of trait lists, rather than a plain string. A correct conversion flattens those attributes into readable personality text.

How do I convert an Agnai character to SillyTavern?

Export the Agnai JSON, then convert it with a tool that maps greeting, sampleChat, and the object persona onto the standard fields before writing a V2 PNG. Verify the greeting and examples afterward.

Is it easier to import into Agnai or export out of it?

Importing into Agnai is more forgiving because it accepts a plain-text persona. Exporting out requires mapping Agnai’s object persona and nonstandard keys onto the standard schema.